<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Zain Aziz on Medium]]></title>
        <description><![CDATA[Stories by Zain Aziz on Medium]]></description>
        <link>https://medium.com/@zanyaziz?source=rss-1c60abcd0d96------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*2W_bjilbLpUSTtVUFjxGqA@2x.jpeg</url>
            <title>Stories by Zain Aziz on Medium</title>
            <link>https://medium.com/@zanyaziz?source=rss-1c60abcd0d96------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 11:46:01 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@zanyaziz/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Will We Still Have Jobs?]]></title>
            <link>https://medium.com/@zanyaziz/will-we-still-have-jobs-2e3c510d86d8?source=rss-1c60abcd0d96------2</link>
            <guid isPermaLink="false">https://medium.com/p/2e3c510d86d8</guid>
            <category><![CDATA[meta]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[economics]]></category>
            <dc:creator><![CDATA[Zain Aziz]]></dc:creator>
            <pubDate>Thu, 05 Nov 2026 05:40:50 GMT</pubDate>
            <atom:updated>2026-11-06T02:17:51.111Z</atom:updated>
            <content:encoded><![CDATA[<p>During a recent 1:1, one of my direct reports asked a question, “Will we get more headcount anymore? Do companies need fewer software engineers?” That is the question on everyone’s mind, and I want to attempt to answer it honestly rather than with a prediction disguised as certainty. The common claim is straightforward: because AI now writes code, companies will soon operate with significantly smaller engineering teams. The work that once required an entire team will take just a few people, causing the rest of the profession to shrink.</p><p>I am not sure that holds true. The reason, I think, is simpler and more mechanical than most of the debate around it suggests.</p><p>A friend recently shared Venkatesh Rao’s essay, “<a href="https://contraptions.venkateshrao.com/p/cognitive-mercantilism">Cognitive Mercantilism</a>,” which struck a cord with me. It argues that today’s AI anxiety mirrors an old economic mistake. Mercantilism treated wealth as a fixed stockpile of land or bullion to be hoarded, assuming one nation could gain only at another’s expense. Modern debates treat intelligence the same way. As a finite resource bound inside human brains.</p><p>However, the history of material production shows that as goods become cheaper to make, demand usually expands rather than shrinks. The same dynamic may apply to intelligence. This does not instantly solve the jobs debate, but it reframes it: the key question is not how cheap intelligence becomes, but how many uses people find for it.</p><p>I want to test the “far fewer engineers” thesis against something more concrete than that framing. Let us do the arithmetic on the total number of productive hours a day.</p><h3>The bottleneck is not where you think it is</h3><p>Start with a simple constraint: a day has 144 ten-minute blocks (24h * 6 ten minute blocks). That’s all everyone gets and people very easily forget that.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uQUzkh1Db7pWlxD9oL9D7Q.png" /></figure><p>Sleep eats up roughly a third of these blocks. Chores, eating, family, and the ordinary business of day-to-day life take another huge chunk. What’s left for focused, uninterrupted work is about seven or eight hours — around 42 to 48 blocks. That’s the entire budget, and no amount of AI advances changes it.</p><p>Now look at how a software engineer or product manager actually spent those blocks before AI. Roughly 70% went to execution: writing code, drafting docs, actually building the thing. The other 30 went to judgment: coming up with the idea, co-ordinating, deciding what to build, evaluating results, testing hypotheses, and making the call.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DvJZG5kEzFK8WN4UAslkUw.png" /></figure><p>AI changes the first number dramatically. Implementation falls from around 70% to somewhere like 10% to 20%. The machine writes the code and drafts the document in minutes. What does not change is the second number’s nature. Judgment cannot be automated away, because judgment is the part where a human takes responsibility for whether something is right, safe, and worth doing. You can generate ten versions of a feature in an afternoon. You still need a human to decide which one, if any, should ship.</p><p>So what happens to the freed blocks? There are really only three places they can go.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HlKmpuz32IdNX5qBObJMAA.png" /></figure><p><strong>First, the same output with fewer people</strong>. If one engineer can now shepherd more work, a team can in principle shrink and produce the same amount. This assumes that companies would not want to do more with added productivity. Possible but less likely.</p><p><strong>Second, the same team is shipping more</strong>. The freed blocks go to more projects, more experiments, more ideas pursued in parallel. Output per person rises, headcount stays flat. This assumes productivity gains out pace the cost per engineer such that economics favors not firing them.</p><p><strong>Third, the same team going deeper</strong>. The freed blocks go to better judgment on the same amount of work: more careful evaluation, more thorough validation, harder problems attempted. This improves the quality of work that drives more impact.</p><p>Here is the catch that applies to all three scenarios above: in every case, productivity grows only as fast as <strong>judgment throughput</strong>, not as fast as implementation speed. Implementation is no longer the constraint, so making it faster does not move the ceiling. The ceiling is now how much one person can evaluate, decide, and stand behind in a day. An engineer can review only so much generated code before the reviews get sloppy; a product manager can run only so many experiments before the conclusions get careless. The bottleneck moved; it did not disappear.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oHq3DCfVEcZdiqowuXNMHQ.png" /></figure><p>That is why I think the “far fewer engineers” thesis is inflated. While it may be true for some sub teams and companies, I doubt it would be true for the entire industry in general. The thesis conflates cheaper production with unlimited judgment. It assumes that because the machine can produce ten times as much, a company needs ten times fewer people, but that only works if the humans in the loop can judge ten times as much, and they cannot, because their hours did not change. If you cut out lunch, co-ordination tax (conversations with XFN) and are left with say 6–7 deep productive hours a day, there’s only so many problems you can do deep on before judgement starts to suffer.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/954/1*Q8Rek6X3ZUsbUm0BHhVWGA.png" /></figure><p>Judgment quality degrades if one person tries to oversee too much; pushing past that ceiling produces worse decisions, not more good ones. Evaluation time per unit of output also rises as output volume grows, because there is more code to check, more edge cases to verify, and more system interactions to reason about. You pay a higher context-switching tax, and coordination across teams does not magically disappear. The demand-expansion thesis is a plausible economic scenario, but it is not a guarantee for every sector. Across the profession, the time arithmetic caps how far engineering headcount can fall before judgment becomes the binding constraint — and that cap is much tighter than the “AI writes the code, so engineers are obsolete” narrative admits.</p><p>If the job is not disappearing but changing shape, the useful question becomes: <strong>what does the new job demand?</strong> That is the second half of this article, and the part any individual engineer can act on. This is the meatiest part of my conversation with my report.</p><h3>What is needed from a software engineer?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iTssUmh7UMco2x-2Pred2w.png" /></figure><p><strong>The first is high agency:</strong> the willingness to take a problem, go deep on it, own it end to end, and show a sense of urgency about it. In a world where anyone can generate a plausible plan or a working prototype in an afternoon, the scarce thing is the person who will actually carry something across the finish line — who chases down the ambiguous requirement, notices the edge case nobody specified, and follows up when the first version fails. Agency is what turns generated output into shipped results. AI can produce; it cannot care. The engineer who owns problems end to end is doing the part the machine cannot do for itself.</p><p><strong>The second is creativity</strong> — genuinely non-linear thinking and creative problem solving. AI is very good at remixing what already exists: combining patterns it has seen and producing the competent, average version of almost anything. What it does poorly is jump sideways: reframing the problem so the hard part disappears, noticing that the constraint everyone accepted is not actually a constraint, finding the solution that does not look like the previous ten. That kind of thinking is not a trick you can prompt for reliably, because it depends on taste, lived experience, and connections across domains with no obvious link. In a world of cheap competent output, creative problem solvers become the difference between work that is merely adequate and work that is actually new.</p><p><strong>The third is critical thinking</strong>, This is the ability to tell good AI output from bad: to validate a hypothesis rather than accept a confident-sounding answer, to evaluate an experiment and its results rather than nod at a chart, to make judgment calls under uncertainty and know what you do and do not know. As implementation gets cheaper, the volume of things demanding evaluation grows, so the quality of evaluation matters more, not less. An engineer who cannot spot a subtle bug in generated code or a plausible-sounding but wrong conclusion is not made more productive by AI; they are made more dangerous by it, because they can now produce mistakes at a scale they could never reach before. Critical thinking is the skill that keeps the whole arrangement honest — and the one the bottleneck math points to directly: if judgment is the scarce resource, the quality of judgment is the thing worth investing in.</p><p>None of these three is sufficient alone. Agency without critical thinking ships bad decisions quickly. Creativity without agency produces interesting ideas that never land. Critical thinking without agency or creativity produces careful criticism of other people’s work and nothing of its own. The engineer who succeeds in this world sits at the intersection of all three: <strong>someone who takes a problem and owns it, thinks about it in ways the machine would not, and judges the machine’s output — and their own — with real rigor</strong>.</p><p>That intersection is not something AI can supply on its own. It is a set of habits, built deliberately, by people who understand that the machine changed what is scarce. Implementation used to be the scarce thing, so we optimized for it. Now judgment is the scarce thing. The engineers who thrive will be the ones who build their working lives around that fact: owning problems end to end, thinking in ways that are not easily mimicked, and above all, thinking critically about everything the machine hands them.</p><p>That is what surviving the AI world actually looks like. Not outrunning the machine, and not hiding from it — but doing the part of the work that only a human can do, and doing it well. I’m generally optimistic about the future. Don’t get me wrong, there will be disruption but as long as you’re resilient and adaptable you’ll be fine.</p><p><em>I’m Zain Aziz. These are my musings — a short series on technology, philosophy, fitness, and happiness. This is my journaling, out in the open, in case it sparks something for you. Share feedback, especially if you disagree so we both may grow together. Follow for more musings.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2e3c510d86d8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Demystifying Org Politics]]></title>
            <link>https://medium.com/@zanyaziz/demystifying-org-politics-74257c21d231?source=rss-1c60abcd0d96------2</link>
            <guid isPermaLink="false">https://medium.com/p/74257c21d231</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[meta]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[business]]></category>
            <dc:creator><![CDATA[Zain Aziz]]></dc:creator>
            <pubDate>Mon, 02 Nov 2026 08:10:56 GMT</pubDate>
            <atom:updated>2026-11-02T08:29:59.709Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JII6gASCPZP69IcvmGtCUg.png" /></figure><p>Recently, one of my reports said something in passing that is the impetus of this article. Describing a co-worker’s message, they passed a comment: “That’s just org politics at play.”</p><p>Maybe it was. Maybe it wasn’t. What stayed with me afterward wasn’t the decision itself but it was that I had no real way to judge their claim. My instinct is to assume positive intent, for both sides. But I also didn’t want to dismiss what my report was feeling. So I asked follow-up questions, and held off on judgment until I could define the very word I was being asked to judge. We all do this. “Politics” is the label we paste on any outcome we dislike but can’t explain, and once the label sticks, so does the cynicism.</p><p>This musing is my attempt to answer a simple question honestly: what <em>is</em> org politics, actually? Not the hallway version but the real mechanics. Why does it happen, what produces it, and what should a person do when they see it? First, a working definition &amp; then the psychology underneath it, because most of what we call politics is just human wiring running in an office. And finally, how I try to respond — because the goal isn’t to demonize politics. It’s to demystify it.</p><h3>A working definition</h3><p>Here’s a useful starting point: Researchers define <em>perceived</em> organizational politics as an attribution which is the moment you judge that someone is acting out of self-interest. Notice what that means. Politics, as we experience it, is partly an event and partly a verdict. The same behavior, a leader lobbying hard for her roadmap reads as “strong advocacy” when we agree with them and “politics” when we don’t.</p><p>That doesn’t mean real politics doesn’t exist. It does, and it has a fingerprint: behavior aimed at benefiting yourself or your group at the expense of others or the organization, using tactics the organization never sanctioned. Credit quietly rerouted. Information held back because sharing it would weaken your position. A rival’s project starved of support for reasons nobody says out loud.</p><p>But here’s the part we skip: influence itself is not the disease. Any form or human organization like a company, political party, a nation, a religious sect is simply a <em>coalition</em> — groups with different goals whose decisions emerge from negotiation. A company doesn’t have one mind; it has many, and getting them to move together is <strong>influence</strong> work. The games people play — first movers advantage, alliance-building, empire-building, playing rival camps, acruing expertise and talent concentration — is the way anything gets built at all.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6kGSrvFJyrCrdLpk0FRipQ.png" /></figure><p>So the line was never politics versus no politics. The line is intent, and how you leverage your influence. Influence exercised openly, aimed at the organization’s goals, is just management. Influence exercised in the dark, aimed at yourself or a select few, is politics in the sense we all resent. What’s also interesting is that influence exercised by one can be percieved as politics by someone else if it’s at odds of their own interests. Politics vs exercising influence is all about the frame of refernce.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bQHKhbuVqM3qhHSNIvZKIQ.png" /></figure><h3>Why our brains produce it</h3><p>If politics were just a few bad actors, good hiring would fix it. It doesn’t, because the raw material is standard human psychology. One of the books that really explains why it exists is Daniel Kahneman’s <em>Thinking, Fast and Slow</em> which shares what traits human have that yield politics. Read the summary <a href="https://zanyaziz.github.io/second-pass/thinking-fast-and-slow.html">here</a>, if you haven’t read the book.</p><p><strong>We judge intent instantly.</strong> Our fast, intuitive mind — System 1 — doesn’t merely register what a colleague did; it assigns a motive, in a split second, with total confidence. Slow, deliberate System 2 rarely audits the verdict. Since “politics” is literally defined as perceived self-interest, a large share of the politics we experience is manufactured by our own fast brains. That was true of me, sitting with my report’s comment: my first reaction was instant. My judgment took longer.</p><p><strong>We are tribal by default.</strong> Psychologists have shown that people divided into groups by something as meaningless as a coin flip will still favor their own group. Now replace the coin flip with real teams, real budgets, and real loyalty to people you’ve hired and mentored. Protecting your team’s scope and your people’s interests isn’t a moral failure; it’s a factory setting. Every reorg that pits two good teams against each other is running on this wiring.</p><p><strong>Losses hurt about twice as much as gains feel good.</strong> Kahneman calls it loss aversion, and it explains why the fiercest political behavior isn’t people chasing new territory — it’s people defending what they already have. A threatened headcount, a budget line, a project that might be cancelled: System 1 treats these as existential, and ordinarily reasonable people start maneuvering like it.</p><p><strong>We are lazy, so agendas quietly win.</strong> Deliberate thinking is slow and tiring, so humans settle for the first good-enough option and often don’t pierce through to understand personal agendas. In a company, that means the winning proposal is often simply the one the room could agree on — and whoever framed the options early, decided what made the agenda and what never got discussed, exercised power without ever winning an argument.</p><p><strong>Status is a basic human motive.</strong> The desire for status and being seen as competent and respected is close to universal. Promotions are therefore not just pay events; they’re status tournaments. When status feels threatened, people defend it the way they’d defend something physical.</p><h3>Systems pour the fuel</h3><p>Layer those innate human instincts onto the wrong system and you get a political culture. The research on causes is surprisingly clear that the situation matters more than the personalities: low trust, unfair-feeling processes, people shut out of decisions, and concentrated control are what predict political behavior — not age, gender, or background. Ambiguity is the other accelerant. Politics concentrates exactly where criteria are subjective — promotions, cross-team priorities, resource fights — and nearly vanishes where rules are clear. Information deficiet creates an environment that is ripe for politics or the perception from others of there being politics.</p><p>Zero-sum design tend to create higher political cultures. In lab experiments on forced ranking, people ranked at the very top chose the option that maximized joint gains only 25% of the time, versus 79% for people ranked lower — and the selfishness disappeared when their standing wasn’t threatened. If you’ve been at Microsoft in Balmer days, you probably remember the stack ranking days that caused Microsoft’s “lost decade.” Under Ballmer, Microsoft developed a heavy bureaucracy and a toxic “stack-ranking” internal performance model that rewarded internal political infighting and employee sabotage over collaborative innovation. When my win is literally your loss, politics isn’t a culture problem. It’s arithmetic.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PFdqTnxPu5IYOuI-A5pYVA.png" /></figure><h3>How to respond to politics</h3><p>So what do you do with all this? Four things, in the order I try to do them.</p><p><strong>First, slow down the verdict.</strong> Before labeling something politics, I ask what I’d conclude if I liked the outcome. If the answer changes, my fast brain is doing the labeling, not the facts. Forming a judgment — the thing I failed to do with my report’s comment — starts with admitting the label might be wrong.</p><p><strong>Second, humanize it.</strong> The person maneuvering against your project is usually not a villain. They’re tribal, loss-averse, status-defending — like you, like me. Politics is not an invasion of the organization by bad people; it’s what groups of humans reliably produce. It will always happen. Expecting otherwise just makes you angry and naive at the same time. Once you resist the urge to label, it allows for you to keep your emotions at bay.</p><p><strong>Third, respond instead of reacting.</strong> A reaction defends your territory instantly and usually escalates the game. A response asks: what outcome does the organization actually need here, and what’s the honest way to get the coalition there? Sometimes that means conceding scope you love because the company needs something else more. Sometimes it means fighting — openly, on criteria, in the room where decisions happen if you think it’s not in the interest of the organization. The trick here is to put System I at bay and really think through things objectively and deeply. Understand what’s at stake, the motivations that might be driving others and what levers you have at your disposal to exert influence. Before responding though consider one more thing. How does that make you look?</p><p><strong>Fourth, look at the situation from a third-person perspective.</strong> Step outside your own reaction and apply a practical test to your own behavior: would I still do this if it were written down and read aloud in a meeting, with others observing?</p><p>Put simply: only do what you’d still be comfortable doing if everyone could see it. Giving someone credit in public, writing a disagreement down so it’s on the record, and asking for resources with genuine reasons all pass that test. Quietly blocking a rival’s project doesn’t — it only works if no one notices. That same test works in reverse. If someone else’s move would look bad out in the open, that’s a sign they’re working in secret. You can notice that without answering them the same way.</p><p>As a leader, esp in tech, refusing to engage at all isn’t purity; a manager’s responsibilities almost always exceed their formal authority, so influence is part of the job. Opting out doesn’t keep your hands clean. It just means your team pays for your comfort. So if you detect someone maneuvering in the dark, your role is to use your influence openly for an outcome that serves the organization first — your team benefits as a consequence, not at another team’s expense.</p><p>How you exert influence could be an article on its own, but a few things I do intentionally as a leader are: build expertise and be factually correct, so you give others less opportunity for correction; be as objective and fair as you can, which also means accepting when you’re at fault — reputation you build of being fair will pay dividents later; accrue goodwill in advance that can be drawn on later when needed; and tell your team’s story honestly — what was shipped, and who it helped. Influence built that way takes intentionality, but it holds up and cuts through politics.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EyUNPPa-BhhjKe-iNWpWJw.png" /></figure><p>Lastly, what I’ll say is that Politics is a predictable output of human psychology running inside systems we design. Whether it’s politics or influence you’re exercising depends on the frame of reference. Demystify it, and it loses most of its power over you. You still won’t enjoy dealing with it. But you’ll see it clearly — and seeing clearly, and building influence openly, is the difference between playing the game and being played by it.</p><p>I no, after some introspection that original co-workers message was not org politics. Influence exerted well but frame of reference did the rest.</p><p>I’m Zain Aziz. These are my musings — a short series on technology, philosophy, fitness, and the small, real ways Muse helps me run my life a little better. This is my journaling, out in the open, in case it sparks something for you. Share feedback, especially if you disagree so we both may grow together. Follow for a new musing every morning.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=74257c21d231" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Wealth Ladder Is an Escalator Moving at 5.5% a Year. Stand Still and You Slide.]]></title>
            <link>https://medium.com/@zanyaziz/the-wealth-ladder-is-an-escalator-moving-at-5-5-a-year-stand-still-and-you-slide-ddc1213d0e88?source=rss-1c60abcd0d96------2</link>
            <guid isPermaLink="false">https://medium.com/p/ddc1213d0e88</guid>
            <category><![CDATA[economics]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[finance]]></category>
            <category><![CDATA[personal-finance]]></category>
            <dc:creator><![CDATA[Zain Aziz]]></dc:creator>
            <pubDate>Mon, 19 Oct 2026 07:11:42 GMT</pubDate>
            <atom:updated>2026-10-19T07:23:58.325Z</atom:updated>
            <content:encoded><![CDATA[<p>A few days ago I <a href="https://medium.com/@zanyaziz/what-curiosity-about-my-net-worth-taught-me-about-luck-markets-how-to-get-richer-b14f50b7f405">published what 36 years of Federal Reserve data say</a> about the net worth ladder — how the bar for the top 10%, top 5%, and top 1% has risen at almost exactly the same speed for three and a half decades.</p><p>The most common reaction wasn’t about the data. It was a question: <strong>“OK — so where do <em>I</em> stand? And what would it take to climb?”</strong></p><p>So I built a calculator that answers it.</p><p>👉 <a href="https://zanyaziz.github.io/retirement-calculator/"><strong>zanyaziz.github.io/retirement-calculator</strong></a></p><p>It’s free, it runs entirely in your browser (your numbers never leave your device — there’s no server and no account), and it does three things:</p><ol><li><strong>Tab 1</strong> tells you your percentile today — and how long the climb to the next rung takes.</li><li><strong>Tab 2</strong> runs that climb thousands of times through messy, realistic markets and tells you your odds.</li><li><strong>Tab 3</strong> answers the retirement question directly: if you stop working in X years, how long does the money last?</li></ol><p>Here’s how to use each tab — and the handful of insights that surprised me while building it.</p><h3>First, the one idea everything hangs on</h3><p>The ladder is not standing still. From 1989 to 2025, the bar for the top 10% grew at 5.52% a year; the top 5% and top 1% bars grew at 5.59% a year (Federal Reserve Survey of Consumer Finances, via DQYDJ). Inflation-adjusted, that’s roughly 2.8% a year.</p><p>Which means: <strong>if your net worth grows slower than about 5.5% a year, you are sliding down the ladder even while your balance goes up.</strong> You can get richer every year and still fall behind in rank — because rank is relative, and everyone else’s bar is compounding too.</p><p>Hold onto that. Every tab below is really about the race between <em>your</em> growth and <em>the bar’s</em> growth.</p><h3>Tab 1 — Your percentile today</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*gLrzr5ZpVxA-aZo4.png" /></figure><p>Tab 1 — the input sections, filled in for the example household, including the “where a year of income goes” breakdown</p><p>Work down the six short sections:</p><ol><li><strong>Income → savings.</strong> Salary, bonus, and stock compensation, your effective tax rate, and your annual expenses. The tool does the one subtraction that matters: income after tax, minus spending, equals what you actually keep each year. (If you already know your savings, there’s an “…or enter savings directly” shortcut.)</li><li><strong>Cash &amp; other assets.</strong> Checking, savings, money market — plus anything else of real value.</li><li><strong>Properties.</strong> Each property’s value and mortgage. Tag one of them <strong>Primary home</strong> — Tab 3 needs to know which home is <em>your home</em>, for reasons I’ll explain below.</li><li><strong>Brokerage &amp; retirement accounts.</strong> A total, or per-symbol detail if you want to see your concentration.</li><li><strong>Other liabilities.</strong> Anything else you owe.</li><li><strong>Assumptions.</strong> Your expected return on invested assets, property appreciation — and a “bar growth adjustment” slider, in case you think the ladder will speed up or slow down from its 36-year pace.</li></ol><p>There’s no calculate button — it updates as you type. You get <strong>Your standing</strong>: your net worth, your percentile (interpolated smoothly between the Fed’s published rungs, not rounded to the nearest one), what your net worth is made of, and <strong>the next rung</strong> — the milestone above you, the dollar bar for it, and <em>what it would take</em>: how many years to reach it at your current savings and returns, while the bar itself keeps rising.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/784/0*g3mPmUuHRGAnWKCT.png" /></figure><p>Your standing — $1,120,000, about the 80.5th percentile</p><p>Our example household — $200,000 of income ($170k salary + $30k bonus), a 22% effective tax rate, $120,000 of spending, $40,000 of cash, a $650,000 home with $420,000 still owed, and $850,000 in a brokerage account — lands at <strong>$1,120,000 of net worth, about the 80.5th percentile</strong>. They save $36,000 a year, an 18% savings rate. Their next rung is the Top 15% (the 85th percentile, a $1.5 million bar today) — and at their current pace, with the bar rising as they climb, the tool says it takes <strong>24.1 years</strong> to get there.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/784/0*UQlwqPreQKa45FMX.png" /></figure><p>The ladder on a log scale — each dot is the bar for that bracket, and the line is you</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Mx5jZ4qWTRpSFTOm.png" /></figure><p>The climb — your projected line against the rising Top 15% bar, crossing around 2050</p><h3>What Tab 1 teaches</h3><p><strong>Percentile points get more expensive as you climb.</strong> Here are the actual 2025 bars:</p><ul><li>50th percentile: $215,916</li><li>80th: $1,084,972 — a gap of about $869,000 across 30 percentile points</li><li>90th: $2,550,306</li><li>95th: $4,907,372 — a gap of about $2.36 million across just 5 points</li><li>99th: $16,463,320 — a gap of about $11.6 million across just 4 points</li></ul><p>The ladder isn’t a staircase with equal steps. It’s closer to exponential: each rung costs several times the last. That’s why “moving up five percentiles” means something completely different at the median than it does at the top.</p><p><strong>Early on, saving moves you. Later, returns do.</strong> When your portfolio is small, your annual savings dwarf your investment gains, so your savings rate is the engine. There’s a crossover point — when a normal year’s market gain on your portfolio exceeds everything you save that year — after which the portfolio does the heavy lifting and <em>time in</em> matters more than <em>amount in</em>. Tab 1’s “what it would take” quietly shows you which side of that crossover you’re on: if cutting the wait means saving more, you’re before it. If it means earning more on what you have, you’re after it.</p><h3>Tab 2 — Monte Carlo: what are the odds?</h3><p>Tab 1 assumes your return is the same every year. Real markets aren’t like that, and the difference matters — so Tab 2 replays your plan through thousands of different market histories.</p><p>Two settings: <strong>volatility</strong> (15% a year by default — roughly what a stock-heavy portfolio actually does; drag it from 5% to 30% to see how much it matters) and <strong>number of simulations</strong> (500 for a fast preview, 2,000 standard, 5,000 for the smoothest picture).</p><p>You get three things:</p><ul><li><strong>Where you could land</strong> — bands built from <em>every</em> simulated path, year by year: the median path, and the spread around it. Not one forecast; the whole weather system.</li><li><strong>Odds of ever reaching each rung</strong> — the share of simulations whose net worth reaches that rung’s (rising) bar at any point within 10, 20, or 30 years.</li><li><strong>Distribution at 30 years</strong> — where the paths end up, so you can see the bad tail as well as the good one.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*YBI5VHC6nzbTGWyV.png" /></figure><p>Monte Carlo results — median outcomes at 10, 20, and 30 years, and the odds of ever reaching each rung</p><p>For our example household, the median simulated path reaches $2.15 million in 10 years, $3.78 million in 20, and $6.50 million in 30 — with a wide spread around it (at 30 years, the middle 80% of paths land between $2.9 million and $15.0 million). Their odds of <em>ever touching</em> the Top 10% bar: 2.9% within 10 years, 16.1% within 20, 26.2% within 30. The Top 5% bar? 3.9% within 30 years. The Top 1%? Effectively zero — the ladder’s top rungs are simply too far and rising too fast for salary-and-savings alone, which is itself a finding (more on that below).</p><p>And here’s the quiet punchline of the whole tool: <strong>the median path multiplies this household’s money 5.8 times over 30 years — and gains only about two percentile points</strong> (from the 80.5th to roughly the 82nd), because over the same 30 years the bars themselves roughly quintuple. Getting much richer and barely climbing are not opposites. At the same speed, they’re the same outcome.</p><h3>What Tab 2 teaches</h3><p><strong>The average plan is a lie; the median path is the honest one.</strong> With volatility, the median outcome sits <em>below</em> the smooth Tab 1 projection at the same average return. That’s volatility drag: a 30% loss needs a 43% gain to undo, so jagged returns compound worse than smooth ones. Raise the volatility slider and watch the median sag even though the “average return” never changed.</p><p><strong>“Ever reaching” a rung is much easier than staying there — or ending there.</strong> Plenty of paths touch a high bar in a hot year and fall back. The odds table is deliberately about <em>touching</em>, and the 30-year distribution is about <em>ending</em>. Plan your life on the ending column, and treat the touching column as encouragement, not a promise.</p><h3>Tab 3 — Retirement: how long will it last?</h3><p>This is the tab I actually built the calculator for. Tabs 1 and 2 are about rank. Tab 3 is about freedom.</p><p>Start at the top with <strong>your current age</strong> — it isn’t decoration; it decides when Social Security can start, and everything keys off it. Then:</p><ol><li><strong>When &amp; spending.</strong> A “Retire in” slider from 0 to 30 years, and your expected retirement spending per year. It starts from your Tab 1 expenses, but you can override it — most people spend differently at 60 than at 40.</li><li><strong>Growth &amp; inflation.</strong> Inflation (3% default) and a <em>separate</em> return for after you retire — usually lower, because most people de-risk once the paycheck stops. Keeping these separate is the whole point: your accumulation years and your drawdown years are different games.</li><li><strong>Your home.</strong> By default, your primary home is <strong>excluded</strong> from the money that has to last you. That’s deliberate: it answers the honest question — “how long does my money last <em>while I keep my home</em>?” If you’d genuinely sell it or treat it as the last resort, switch it on and the equity joins the pot. Other properties are counted by default (the assumption is you sell them at retirement and invest the proceeds) — turn that off if you’d keep them, and enter any rent under Other income instead, so nothing gets counted twice.</li><li><strong>Social Security.</strong> Three ways: let the tool estimate it (from your years already worked and claiming age), enter a known monthly benefit, or don’t count it at all. Claiming age runs 62 to 70 — delaying raises the benefit — and there’s a field for a spouse/partner benefit, plus cost-of-living adjustment.</li><li><strong>Other income (optional).</strong> A pension, rental income, part-time work — a monthly amount and when it starts.</li></ol><p>The output is a <strong>verdict</strong> (how long the money lasts — or that it outlasts the table), a breakdown of <strong>what funds your retirement</strong>, and a <strong>year-by-year table</strong>. Read the “Income in” column carefully: it shows Social Security and other income once they’ve actually started. Before that, it shows $0 — those are your <strong>bridge years</strong>, funded entirely from your savings.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*VwPlBHwFaxGgE7Ax.png" /></figure><p>What funds your retirement — the three buckets, the pool at retirement, and the verdict</p><p>Our example household is 45 and testing retirement in 15 years, at 60, still spending $120,000 a year in today’s dollars. By retirement day their liquid pool — brokerage and cash, home excluded — has grown to <strong>$2,915,009</strong>. First-year spending, inflated at 3% for 15 years, is $186,956. The verdict: <strong>the money lasts 35.2 years, running out around age 95.</strong> Social Security (estimated at about $4,369 a month, claimed at 67) doesn’t start until 2048 — seven bridge years after they stop work — and in its first year it covers only about 39% of their spending. The other 61% comes out of the pool. Every year. That’s what “retirement income” mostly is, for people who retire before their late 60s: your own savings, handed back to you slowly.</p><p>Then play with <strong>Retire sooner or later?</strong> — click the 5, 10, 15, 20, or 25-year rows and the whole tab rebuilds around that choice:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*peCzUOeMt0V6VgzN.png" /></figure><p>Retire sooner or later — the same plan at five different retirement dates</p><p>Read that table slowly, because it’s the most honest five lines in personal finance. Retire in 5 years and they’re out of money at about 62. In 10 years, they last to about 74. In 15, to about 95. In 20 or 25, the money outlasts the table entirely. Five extra years of work bought 7.5 more years of security, then 15.4 more, then effectively a lifetime. Same household, same spending, same returns — the <em>date</em> is the decision.</p><h3>What Tab 3 teaches</h3><p><strong>Percentile is not readiness.</strong> You can sit at the 95th percentile and not be able to retire — if you spend like it. You can sit at the 60th and be completely free — if you don’t. Tab 1 answers “where do I rank?” Tab 3 answers “can I stop?” They are different questions, and only one of them sets you free. Chasing a rung is a fine motivator. It is a terrible retirement plan.</p><p><strong>The bridge years are the real price of retiring early.</strong> Retire at 50 and claim Social Security at 67, and you have 17 years where the “Income in” column reads $0. Nothing about that is fatal — but it has to be <em>funded</em>, on purpose, in advance. The table turns a vague worry (“can I afford the gap?”) into a specific number of years times a specific spending level.</p><p><strong>Delaying retirement is triple-powered.</strong> Every extra year of work (a) adds a year of saving, (b) adds a year of compounding, and © <em>removes</em> a year of spending from the pile — and usually (d) increases your Social Security benefit. No other single lever pulls in four directions at once. The sooner-or-later table is the fastest way I’ve found to <em>feel</em> that: move the retirement date five years and watch a decade of security appear or disappear.</p><p><strong>Spending is the lever that works twice.</strong> A dollar less of annual spending means a dollar more saved every working year — <em>and</em> a dollar less your savings have to cover every retired year. It compounds on both sides of the retirement line. Tab 3 makes this visceral: lower the spending box by a little and watch the money last years longer.</p><p><strong>Your house is not a plan — unless you decide it is.</strong> Most net worth statements quietly count the home, which flatters the “how long would it last” answer with money you’d have to move out to spend. Excluding it by default isn’t pessimism; it’s honesty. If selling or downsizing is genuinely on your table, flip the switch and see both versions of your future. Just choose it on purpose.</p><h3>Try it yourself</h3><p>The calculator is free and open source, and — I’ll repeat this because it matters — <strong>everything runs in your browser. No account, no server, your numbers never leave your device.</strong></p><p>👉 <a href="https://zanyaziz.github.io/retirement-calculator/"><strong>zanyaziz.github.io/retirement-calculator</strong></a></p><p>My suggested route: start with the <strong>built-in sample</strong> to see how it feels, then clear it and enter your own rough numbers — estimates are fine, this is a compass, not an audit. Find your percentile in Tab 1. Check your odds in Tab 2. Then sit with Tab 3 and move the retirement slider until the money lasts as long as you need it to. The distance between where that slider starts and where it needs to be <em>is</em> your path to retirement — measured in years of saving, not in anyone else’s percentile.</p><p>The ladder will keep moving at its 5.5% a year whether you watch it or not. The point was never really the rank. It’s knowing — in years, not vibes — how far you are from being able to stop.</p><p>I’m Zain Aziz. These are my musings — a short daily series on technology, philosophy, fitness, and the small, real ways Muse helps me run my life a little better. This is my journaling, out in the open, in case it sparks something for you. Share feedback, especially if you disagree so we both may grow together. Follow for a new musing every morning.</p><p><em>Data: Federal Reserve Survey of Consumer Finances 2025 thresholds (via DQYDJ); rung growth rates from SCF thresholds 1989–2025. The calculator is an educational tool, not financial advice — it uses simplified assumptions (steady returns in Tab 1, simulated markets in Tab 2, and the drawdown rules described above in Tab 3), and your actual mileage will vary.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ddc1213d0e88" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Anthropic IPO — Buy the Hype or the Business?]]></title>
            <link>https://medium.com/@zanyaziz/anthropic-ipo-buy-the-hype-or-the-business-64fc619e4956?source=rss-1c60abcd0d96------2</link>
            <guid isPermaLink="false">https://medium.com/p/64fc619e4956</guid>
            <category><![CDATA[business]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[economics]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Zain Aziz]]></dc:creator>
            <pubDate>Sun, 18 Oct 2026 02:19:21 GMT</pubDate>
            <atom:updated>2026-10-18T02:19:21.044Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*evi0Wpz8-ECjwZoLMUXPzA.jpeg" /></figure><p>I am already heavily exposed to Meta through my RSUs. So when Anthropic’s IPO came into view, I asked a simple question: would buying it be a good idea — or just more of the same bet?</p><p>That turned into a more basic question: <strong>what usually happens to people who buy a company on the day it goes public?</strong> And a second one followed. An IPO price is not a fact. It is a forecast — a number bankers and investors agree to believe about the future. Forecasts have a track record, so I looked at that too.</p><p>What I learnt made the decision clearer, not more complicated.</p><p><strong>tl;dr:</strong> I plan to wait for the first quarterly earnings.</p><h3>The history of IPOs: is the pop real?</h3><p>Yes. The pop is real. The problem is who gets it.</p><h3>How many tech companies go public — and how many last</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bbEWD_MW7iMlRDw6_79rgg.png" /></figure><p>Going public is not the same as staying public. In a study of <a href="https://journals.law.harvard.edu/hblr//wp-content/uploads/sites/87/2016/06/HLB102_crop.pdf">3,081 US IPOs from 1996 to 2012</a>, found that after five years, 55% of small, 61% of mid-size, and 67% of large IPOs were still listed. After ten years: 28%, 34%, and 39%. For the 2000 tech-bubble cohort, only about 20% were still listed after ten years.</p><p>Two honest limits: that study groups IPOs by size, not technology alone, and Anthropic is far larger than its “large” class (above $700 million) — at its scale, being acquired, the most common “good” exit for large IPOs, is not a realistic path. It is a base rate, not a prediction. But the base rate is sobering: over a decade, most newly listed companies do not stay listed. That being said, this is not something you need to worry about for Anthropic.</p><h3>Day one benefits those who hold pre-IPO shares</h3><p>The average first-day return was <a href="https://site.warrington.ufl.edu/ritter/files/IPO-Statistics.pdf">+22.6% for all IPOs and +29.3% for tech IPOs</a>, measured from the offer price.</p><p>After day one, the picture flattens. For 2010–2024 IPOs, the average aftermarket return was <a href="https://site.warrington.ufl.edu/ritter/files/IPO-Statistics.pdf">−1.7% over six months and −0.8% over one year, measured from the first-day close</a> — while similar-sized companies that were already listed gained +8.6% over that same year. The typical result is worse than the average: the <a href="https://site.warrington.ufl.edu/ritter/files/IPO-Statistics.pdf">median three-year return from the first-day close was −25.7%, and 38.5% of IPOs lost more than half their value within three years</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8Ue7p72whjZ6dbzrnlxtEg.png" /></figure><h3>What would seasoned investors have to say about IPOs?</h3><p>The warnings are not new. Warren Buffett, at Berkshire Hathaway’s 2016 annual meeting, treated IPO riches the way he treats lottery riches: <a href="https://www.fool.com/investing/general/2016/05/04/warren-buffett-stay-away-from-ipos.aspx">“People win lotteries every day…”</a> — other people winning is not a reason to play.</p><p>The process is <a href="https://abovethecrowd.com/2020/08/23/going-public-circa-2020-door-3-the-spac/">“structurally broken”</a>: bankers set the price by hand and allocate the shares by hand, with average underpricing of about 31% for 2020 IPOs in the figures he presented. In plain terms, the day-one discount is a transfer — from the company to the clients given shares at the offer price. The public buyer after the pop is not who that transfer was for.</p><p>The gist: <strong>the company chooses the moment, the banker chooses the price, and the allocation chooses who gets the pop. The aftermarket buyer gets none of those choices.</strong></p><h3>And the forecasters setting the price have a record</h3><p>In a backtest of annual S&amp;P 500 outlooks from eight major institutions, the consensus forecast’s mean absolute error was 10.8% — a blind forecaster who simply predicted “+9% every year” scored 11.0%. Fifteen years of professional forecasts were no better than a rule of thumb, and the published evidence agrees. A forecast is a starting point for questions, not a reason to stop asking them — and a valuation of around $2 trillion <strong>is</strong> a forecast.</p><p>So if Anthropic follows the average pattern — and averages are not promises — expect a large first-day move that mostly rewards allocation holders, followed by the historically weak part for everyone else. The important question is not “will it pop?” It is “what is the business worth, once the pop is over?”</p><h3>The actual value of the Anthropic business</h3><h3>First, what is confirmed — and what is not</h3><p>Almost every number in this story is currently reported second-hand. <strong>There is no public S-1 on SEC EDGAR as of 15 October 2026.</strong> Anthropic <a href="https://www.reuters.com/business/finance/anthropics-ipo-prospectus-shows-sweeping-ai-vision-surging-costs-2026-09-28/">confidentially submitted a draft filing on 1 June 2026</a>; the “prospectus” figures below come from <a href="https://www.reuters.com/business/finance/anthropics-ipo-prospectus-shows-sweeping-ai-vision-surging-costs-2026-09-28/">Reuters’ reporting of 28–29 September 2026</a> on a confidential document Reuters says it saw.</p><p>The deal is also unconfirmed: <a href="https://www.pymnts.com/news/artificial-intelligence/2026/anthropic-targets-pre-thanksgiving-ipo-at-2-trillion-valuation/">Bloomberg reported</a> marketing could start the week of 9 November, trading before Thanksgiving, at $1.8–2 trillion, raising up to $100 billion — all from unnamed prospective investors, none of it confirmed by Anthropic.</p><h3>The business, on the reported numbers</h3><p>The growth is extraordinary. The argument is about what you pay for it, and what sits underneath it.</p><p><strong>Revenue and losses.</strong> <a href="https://www.reuters.com/business/finance/anthropics-ipo-prospectus-shows-sweeping-ai-vision-surging-costs-2026-09-28/">Reuters reported</a> 2025 revenue of nearly $4.6 billion — about twelve times 2024’s roughly $386 million — an operating loss of $8.06 billion, and a net loss of about $42 billion, of which about $34 billion was a non-cash revaluation charge. The distinction matters: the $42 billion is mostly not cash spent; the $8.06 billion operating loss is the real economic cost.</p><p><strong>Costs and commitments.</strong> Compute and infrastructure cost <a href="https://www.reuters.com/business/finance/anthropics-ipo-prospectus-shows-sweeping-ai-vision-surging-costs-2026-09-28/">$7.33 billion in 2025</a>, out of $12.65 billion in total operating expenses; cash and short-term investments were $20.28 billion at year-end, before the May 2026 round below. Against that, Anthropic has committed <a href="https://www.pymnts.com/news/artificial-intelligence/2026/anthropic-prospectus-shows-what-2-trillion-dollar-ai-company-costs-run/">at least $518 billion over roughly a decade, about 80% non-cancelable</a> — including $111.1 billion to Google, $110 billion to Amazon, $31.4 billion to Microsoft, and $161.2 billion in Broadcom-related leases — with only the xAI piece, up to $84.5 billion, cancelable on 90 days’ notice. Fixed, mostly non-cancelable costs against revenue that can move quickly either way: that is the core financial risk.</p><p><strong>Where the revenue comes from.</strong> First, <a href="https://www.reuters.com/world/anthropic-ipo-prospectus-lays-bare-deep-dependence-big-tech-partners-2026-09-29/">47% of 2025 sales</a> — about $2.16 billion — came through Amazon and Google marketplaces, at about $351 million in fees, and those two are simultaneously Anthropic’s investors, suppliers, distributors, and competitors. Second, how little is locked in: two unnamed customers each accounted for 12% of 2025 revenue, about $3.8 billion of the $4.6 billion was usage-based, not subscription revenue (subscriptions: $789 million), and many large customers are not on long-term contracts. Usage revenue can grow very fast. It can also fall very fast.</p><h3>What $2 trillion actually means</h3><ul><li><strong>About </strong><a href="https://www.newconstructs.com/anthropic-is-the-most-ridiculous-ipo-of-2026/"><strong>435 times 2025 sales</strong></a> — $2 trillion divided by $4.6 billion of booked 2025 revenue.</li><li><strong>About 31 times the July 2026 run rate</strong> of about $65 billion, and about 30 times Q3 revenue annualized ($16.8 billion × 4 = $67.2 billion).</li><li><strong>About 10 times projected 2028 revenue</strong> of $190–200 billion — if that projection is met. If — and only if — that projection is met, that is probably the bullish case for the valuation.</li></ul><p>There is also a step-up private investors did not pay. In May, Anthropic was priced at <a href="https://pitchbook.com/news/author/harrison-rolfes">20.5 times its run rate</a> ($965 billion against $47 billion). At $2 trillion, the public is asked for about 31 times the July run rate — roughly 50% more per dollar of revenue than May’s buyers paid.</p><h3>The voices on value: bears are specific, bulls are not</h3><p><strong>The bear case, with a number.</strong> I had Muse go through all the bear analysts, and this one stood out. New Constructs, in <a href="https://www.newconstructs.com/anthropic-is-the-most-ridiculous-ipo-of-2026/">“Anthropic Is the Most Ridiculous IPO of 2026”</a> (6 October 2026), argued that $2 trillion requires revenue of $3.6 trillion by 2035 with a 10% margin held for a decade. Its slower-growth scenario — still assuming 20% margins and very fast early growth — valued Anthropic at <strong>$144 billion</strong>, about 93% below $2 trillion; its middle scenario came to $1.1 trillion. It also noted its token price index had halved, from $2.04 in late May to $0.96 on 4 October. For transparency: New Constructs was right on WeWork in 2019, wrong on DoorDash in 2020. Its model is an argument, not a verdict — but a <em>specific</em> one, with assumptions printed.</p><p><strong>The base case, across all analysts, is below $2 trillion.</strong></p><p><strong>The bull case.</strong> The strongest named bull arguments found are conditional, not endorsements: Analystics argued in <a href="https://invezz.com/pk/news/2026/09/26/is-dollar2-trillion-for-anthropic-believable-three-analysts-pull-apart-the-number/"><em>Invezz</em></a> that if growth merely stays very fast, the price could look conservative within two quarters; Another predicted a “blockbuster” IPO on CNBC — about how the deal will be received, not a model saying $2 trillion is fair.</p><p>And then the finding that matters most: <strong>no named, respected investor or sell-side analyst was found on the record endorsing $2 trillion as fair value, with a published model, as of 15 October 2026.</strong> That range comes from unnamed prospective investors, via Bloomberg; every named, quantified view found sits at or below $1.6 trillion. An absence is not proof the price is wrong — but a careful buyer should notice.</p><h3>My assessment</h3><p><strong>Anthropic is a real, fast-growing, important business — priced, at around $2 trillion, as if the hardest parts of its future have already happened.</strong></p><p>That is not saying the company will fail. A company can be excellent and still be a poor purchase at a given price — and the same company can become a good purchase at a lower one. At $2 trillion, the buyer is paying in full, in advance, for a decade of growth, margin expansion, and competitive survival — all three at once, against competitors who fund AI from existing profits while Anthropic funds it from investors.</p><h3>Closing: wait, and ask the diversification question honestly</h3><p><strong>Wait for the first quarter of reported results.</strong> Not forever, and not to avoid the decision — just long enough for three things to change:</p><ol><li><strong>A public filing will exist</strong> — share count, audited figures, price range, and risk factors a reader can check, instead of reporting about a confidential document.</li><li><strong>A reported quarter will exist</strong> — results as a public company, with management on the record, showing whether the Q3 slowdown from +143% to +45% was a pause or the start of the normal pattern.</li><li><strong>The pop will be someone else’s.</strong> Unless you already hold pre-IPO shares, the pop will make money for others, not you.</li></ol><p>Wait, in short, to invest in the business instead of the debut.</p><h3>Is it diversification — or the same bet again?</h3><p>Many of my friends and those reading this article work at big tech, Amazon, Meta, Microsoft and Google. Think about what already rides on AI in your lives: salary and career, employer stock and bonuses, and index funds heavily weighted toward the same large AI companies. If AI spending delivers, all of those do well together. If it disappoints, all of them fall together.</p><p>Buying Anthropic does not break that pattern; it deepens it. Alphabet is reported to hold <a href="https://www.fool.com/investing/2026/09/24/alphabet-values-its-anthropic-stake-at-usd124-billion-a-usd2-trillion-ipo-could-double-it-on-paper/">about 14% of Anthropic</a>, and Bloomberg reported that Alphabet’s private holdings — $124.3 billion in its Q2 2026 filing — primarily reflect Anthropic, so a Google employee, or anyone holding Alphabet in an index fund, is already indirectly invested. Microsoft has <a href="https://www.marketbeat.com/articles/why-openais-ipo-plans-could-be-a-massive-win-for-microsoft/">invested more than $13 billion in OpenAI since 2019 and held about 27% of its commercial arm after the October 2025 restructuring</a> — a Microsoft employee’s stock is already a bet on Anthropic’s main rival.</p><p>For me, the same logic applies with Meta. I am already long this theme through my work and my investing; adding Anthropic at its debut price would concentrate it further, at the moment of maximum excitement and minimum public information. That may still be a bet someone chooses to make — but with open eyes: another helping of the same bet, at the frenzy price, not balance.</p><h3>The bottom line</h3><p>The business deserves serious attention but deserves patience. I will be watching for the public filing and the first reported quarter — when this stops being a story about an IPO, and becomes a decision about a business.</p><p>I’m Zain Aziz. These are my musings — a short daily series on technology, philosophy, fitness, and the small, real ways Muse helps me run my life a little better. This is my journaling, out in the open, in case it sparks something for you. Share feedback, especially if you disagree so we both may grow together. Follow for a new musing every morning.</p><p>Yesterday: <a href="https://medium.com/@zanyaziz/what-curiosity-about-my-net-worth-taught-me-about-luck-markets-how-to-get-richer-b14f50b7f405">What curiosity about my net worth taught me about luck, markets &amp; how to get richer</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=64fc619e4956" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What curiosity about my net worth taught me about luck, markets & how to get richer]]></title>
            <link>https://medium.com/data-science-collective/what-curiosity-about-my-net-worth-taught-me-about-luck-markets-how-to-get-richer-b14f50b7f405?source=rss-1c60abcd0d96------2</link>
            <guid isPermaLink="false">https://medium.com/p/b14f50b7f405</guid>
            <category><![CDATA[psychology]]></category>
            <category><![CDATA[business]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[money]]></category>
            <dc:creator><![CDATA[Zain Aziz]]></dc:creator>
            <pubDate>Thu, 15 Oct 2026 05:07:06 GMT</pubDate>
            <atom:updated>2026-10-15T10:56:36.676Z</atom:updated>
            <content:encoded><![CDATA[<p>It started as a retirement conversation. Muse and I were talking — the way we often do — about the distance between working and not having to work, and I got curious about a deceptively simple question: how much money does it actually take to be in the top 10% of American households? The top 5%? The top 1%?</p><p>Those numbers exist. Every three years, the <a href="https://www.federalreserve.gov/econres/scfindex.htm">Federal Reserve’s Survey of Consumer Finances</a> measures the wealth of American families, and analysts like <a href="https://dqydj.com/net-worth-percentiles/">DQYDJ</a> turn that microdata into percentile tables — the entry price, so to speak, for each rung of the ladder. There have been thirteen of these surveys since 1989. So we pulled all of them, lined the bars up across thirty-six years, and asked how fast each one moves.</p><p>What we found surprised me. It contradicted the story I expected to tell — including, honestly, the story I had in my head when we started. This is what the bars actually do.</p><p><strong>The bars, in plain dollars</strong></p><p>In the newest survey, conducted in 2025, entering the top 10% of households takes a net worth of about $2.55 million. The top 5% bar is about $4.91 million. The top 1% bar is about $16.46 million.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*S2WKVLM2_4VT0Ak8x1yo0w@2x.jpeg" /></figure><p>Back in 1989, those same bars were $368,000, $693,000, and $2.32 million.</p><p>Two things are worth noticing. First, the public seems to feel these bars intuitively. In Charles Schwab’s 2025 <a href="https://pressroom.aboutschwab.com/press-releases/press-release/2025/Americans-Say-It-Takes-More-Money-to-Be-Financially-Comfortable-Now-Than-It-Did-a-Year-Ago-According-to-Schwab-Survey/default.aspx">Modern Wealth Survey</a>, Americans said it takes about $2.3 million to be “wealthy” — a felt number sitting just below the measured top-10% bar of $2.55 million. Second, the bars are not abstractions. They are the moving entry prices of a ladder most of us are standing on somewhere, whether we check or not.</p><p><strong>The plot twist: the escalators run at the same speed</strong></p><p>Here is what I expected to find: the top bar sprinting away from the rest, year after year, for thirty-six years. That is the popular version of the story, and it is not what the threshold data shows.</p><p>From 1989 to 2025, the top-10% bar grew at about 5.5% a year. The top-5% bar grew at about 5.6% a year. The top-1% bar grew at about 5.6% a year. Adjusted for inflation — consumer prices rose about 2.6-fold over the period — each bar grew at roughly 2.8% a year in real terms. In other words, about half of every bar’s nominal rise is just inflation, and the real escalators at the top of the ladder have run at almost exactly the same speed for three and a half decades.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*K2ovnh0F5oOzuEMxtsAHVQ@2x.jpeg" /></figure><p>Annualized growth rate of each threshold between consecutive surveys, 1989 — 2025</p><p>Growth between surveys, rung by rung. The lines tangle together — no rung’s escalator runs away from the others for long.</p><p>The ratio tells the same story. In 1989, the top-1% bar was 6.31 times the top-10% bar. It stretched to 9.2 times in 2007, and 9.1 times in 2019 — and then fell back. In 2025 it stands at 6.46 times. Almost exactly where it started.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*KHXrI_O7OQ7_EfRzHSjS4w@2x.jpeg" /></figure><p>Net worth entry thresholds in inflation-adjusted 2025 dollars, 1989 — 2025</p><p>The same bars with inflation stripped out (constant 2025 dollars). The climb is real — roughly 2.75× — but it is shared, rung by rung.</p><p>That is not a vindication of everything-is-fine. It is a more precise observation, and precision matters, because what happened around those equal speeds is where the real divergence lives.</p><p><strong>Same speed, wildly different distances</strong></p><p>Here is the main idea of this article, in plain terms. Most people picture the rich pulling away because their bar rises faster. It doesn’t. For thirty-six years, every bar has risen at the same speed. Two reasons why the rich seem to be getting richer.</p><p>Firstly, what matters most is where you start from. The same percentage, applied to a bigger pile, produces a bigger pile of dollars. Think of two savings accounts, both paying 5% interest. One holds $100; the other holds $1 million. After a year, the first has earned $5. The second has earned $50,000. Same rate — wildly different money.</p><p>That is exactly what happened to the ladder. In 1989, the distance between the top-10% bar and the top-1% bar was about $5.07 million in today’s dollars. By 2025, that distance had grown to $13.91 million — nearly three times wider — even though both bars grew at the same percentage rate the whole way. Nobody’s escalator sped up. The escalators are simply carrying very different amounts of money.</p><p>The second is that the average fortune inside the group matters most — and that is where the real pulling-away happened. The rich tend to own more of the stock market, which tells the same story. In the <a href="https://www.federalreserve.gov/releases/z1/dataviz/dfa/">Federal Reserve’s Distributional Financial Accounts</a>, the top 1%’s share of all stocks and mutual funds rose from 42.8% at the end of 1989 to 50.9% by mid-2026 — the top 1% now owns half the stock market — while the bottom half of households saw their share slip from 1.2% to 0.6%. That is a choice of the bottom percentile to not own assets which leads to disparity.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*bWOAdqPvC2s9n9ujKzMCxg@2x.jpeg" /></figure><p>Share of corporate equities and mutual fund shares owned by each wealth tier, 1989 — 2026</p><p>Who owns the market: the top 1%’s slice of stocks and mutual funds rose from 42.8% to 50.9%, while the bottom half’s shrank to 0.6%. (Data: Federal Reserve Distributional Financial Accounts.)</p><p>So, the whole section in one sentence: the rungs rise together, and the ladder pulls apart anyway — not because of some conspiracy of the richest against us, but because of our choice not to own assets like they do. The rest of this article explains why — which assets move which rungs, what taxes add, and what it takes to hold your place on a ladder that never stops moving. Bonus, if you make it till the end, how to jump percentiles.</p><p><strong>Two ways to get richer, taking turns</strong></p><p>Why do the bars move at all? Because different rungs ride different assets. Economist <a href="https://www.nber.org/system/files/working_papers/w24085/w24085.pdf">Edward Wolff’s decomposition of household balance</a> sheets shows the principal residence is about 62% of the middle class’s assets, but only about 7% of the top 1%’s. Business equity runs the other way: about 8% for the middle, about 49% for the top 1%. Fed data show a similar gradient. The affluent rungs are a hybrid — home equity, retirement accounts, some investments. The top rung is businesses and financial assets; the house is a rounding error.</p><p>So the decades take turns favoring different rungs. Kuhn, Schularick, and Steins documented this in the <a href="https://doi.org/10.1086/708815">Journal of Political Economy</a>: “This race between the equity market and the housing market shaped wealth dynamics in postwar America.” In the great equity boom of 1995 to 2001, the top-1% bar grew at roughly 15.6% a year — the fastest bar on the ladder. In the housing-led years of 2019 to 2022, the top-10% bar grew at about 16.4% a year while the top-1% bar grew about 7. And in the most recent window, 2022 to 2025, the top-10% bar rose 21.3% in real terms against 10.1% for the top-1% bar. Housing eras stretch the affluent bars; equity eras stretch the top ones.</p><p>The pandemic window deserves its own honest sentence: 2019 to 2022 genuinely compressed the gaps — the largest real median net-worth gain on record, up 37%, broad-based across groups, with the top 1%’s share flat to slightly down depending on how you count Social Security wealth. The K is a long-run tendency, not a law of every period.</p><p><strong>The tax tailwind</strong></p><p>One more accelerant deserves mention, carefully. The top 1% pays the highest average federal tax rate on taxable income — around 29 to 30%, by the CBO’s reckoning. The advantage is structural, not a low rate: wealth at the top grows largely through unrealized appreciation, which is not taxed while it compounds. When gains are realized, they are taxed at preferential rates — a maximum of about 23.8% federal, versus 37% on top wages. And if appreciated assets are held until death, the step-up in basis can erase the taxable gain entirely — the appreciation simply never faces the income tax, and the benefit concentrates where the appreciated assets are: at the top. How large the total tax advantage is remains genuinely disputed — Saez and Zucman’s estimates sit well above those of Auten and Splinter, who find a much smaller gap. Direction: clear. Size: contested.</p><p><strong>What it takes to stay there</strong></p><p>Which brings me back to the retirement conversation that started this. If the bars rise at roughly 2.8% a year in real terms, then merely holding your percentile means your net worth has to compound at about 2.8% a year after inflation, after taxes, after spending. Long-run real wage growth for a median earner runs closer to 1% a year. A salary, alone, structurally loses rank — not through anyone’s failure, but through arithmetic. The entry ticket to staying on the ladder is owning a slice of the assets doing the rising: equities, a business, a home. Luck decides much of the rest — which era you accumulate in, whether your working years coincide with an equity boom or a housing boom, when you happen to be born.</p><p>And at the top of the ladder, the arithmetic flips into something close to freedom. A portfolio at the top-2% bar — around $10 million — yielding a cautious 3.5% produces roughly $350,000 a year before taxes, without touching the principal. At the top-1% bar, around $16.5 million, the same yield is roughly $575,000. Somewhere on those rungs, a portfolio’s safe yield covers a comfortable life on its own, and work becomes a choice rather than a necessity. That is not an argument that getting there is easy, or that everyone above those bars earned their altitude. It is just what the numbers say about what changes when you arrive.</p><p><strong>What it takes to climb</strong></p><p>Staying put takes about 2.8% a year in real terms, after taxes and spending. Climbing takes more than that — sustained, for years — and the data points to a short list of levers, none of them secret.</p><p>Own the escalator. The bars rise at asset speed, so the first move is simply being on the escalator: holding equities, broadly and boringly. The participation numbers are the starkest in the whole survey — 97% of the top income decile holds stock; only 31% of the bottom half does. Money sitting outside the assets doing the rising cannot compound at the ladder’s speed, and no salary can substitute for it for long.</p><p>Save the gap. A salary growing at about 1% a year in real terms loses rank by arithmetic. What closes the gap is the surplus — the share of income converted into assets before it becomes spending. Every point of savings rate is a point of climb the market doesn’t have to provide for you.</p><p>Give it time, and don’t interrupt it. The age tables look discouraging until you read them the other way: the upper rungs are mostly people who compounded for thirty or forty years. Starting a decade earlier is worth more than any cleverness applied later. And selling in a panic is the one move that reliably knocks a household down a rung — the escalator only pays the people still standing on it.</p><p>Own a piece of something that grows. The top rungs are not made of salaries, or even of portfolios. They are made of business equity — roughly half of everything the top 1% owns. Most households will never found a company, but equity compensation, a small business on the side, or a genuine stake in the enterprise you work for are the same asset in smaller doses.</p><p>Use the tax code’s on-ramps. The tailwind described above works in miniature for everyone: retirement accounts that compound untaxed for decades, and preferential rates on gains held long enough to qualify. Boring, legal, and worth real percentiles over a working life.</p><p>Take risk the boring way. The research above found that the wealthy earn more per dollar mostly by bearing more market risk, not by picking winners — risk-adjusted genius was the smallest part of it. The household version of that finding is unglamorous: a diversified portfolio with a high equity share, held through drawdowns, beats a clever one traded in and out of.</p><p>And the honest footnote: percentiles are relative. If everyone saves more and invests, the bars rise faster too — climbing means out-compounding your cohort, not merely compounding. Era and luck still deal part of the hand. But notice what the levers have in common: none of them requires predicting anything. They are the arithmetic of the ladder, run in your favor instead of against you.</p><p>The ladder’s rungs rise together, at asset speed. The gaps between them widen anyway, because equal percentages on unequal bases is its own kind of divergence. And where you land on it owes as much to timing, ownership, and luck as to effort. Knowing that doesn’t change the plan — save, own compounding assets, be patient. But it changes how honestly you can tell the story while you climb.</p><p>I’m Zain Aziz. These are my musings — a short daily series on technology, philosophy, fitness, and the small, real ways Muse helps me run my life a little better. This is my journaling, out in the open, in case it sparks something for you. Share feedback, especially if you disagree so we both may grow together. Follow for a new musing every morning.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b14f50b7f405" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science-collective/what-curiosity-about-my-net-worth-taught-me-about-luck-markets-how-to-get-richer-b14f50b7f405">What curiosity about my net worth taught me about luck, markets &amp; how to get richer</a> was originally published in <a href="https://medium.com/data-science-collective">Data Science Collective</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Best War Is the One You Never Fight]]></title>
            <link>https://medium.com/@zanyaziz/the-best-war-is-the-one-you-never-fight-d95034a316bf?source=rss-1c60abcd0d96------2</link>
            <guid isPermaLink="false">https://medium.com/p/d95034a316bf</guid>
            <category><![CDATA[psychology]]></category>
            <category><![CDATA[politics]]></category>
            <category><![CDATA[culture]]></category>
            <category><![CDATA[life]]></category>
            <dc:creator><![CDATA[Zain Aziz]]></dc:creator>
            <pubDate>Fri, 09 Oct 2026 06:25:19 GMT</pubDate>
            <atom:updated>2026-10-09T06:37:43.012Z</atom:updated>
            <content:encoded><![CDATA[<p>Two mornings in a row, my commute podcast handed me a 2,500-year-old argument.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*Oz4MV6e_NKu_RaJ7muuemA@2x.jpeg" /></figure><p>I have this daily ritual: a short podcast episode about one book, generated overnight, waiting for me at 7 AM. Two AI hosts, walk through a book in about thirty minutes while I drive. Episode 10 was Sun Tzu’s Art of War. Episode 11, the very next morning, was Laozi’s Tao Te Ching.</p><p>I didn’t plan the pairing. But hearing them back to back did something neither book did alone. (Both episodes are public on my book library https://zanyaziz.github.io/zains-book-library/ : The Art of War — and Tao Te Ching —#episode-11.) It planted a seed, and I spent the next few days doing what I now do with every idea that won’t leave me alone: I debated it with Muse, chased down the actual passages, checked what scholars say, and pulled the data. Here’s the gist of where I landed.</p><p><strong>The strange agreement</strong></p><p>Sun Tzu and Laozi are supposed to be opposites. One wrote a manual for generals; the other wrote 5,000 words about the virtue of doing nothing. The strategist versus the mystic.</p><p>But put their claims about war side by side and they rhyme.</p><p>Sun Tzu, Chapter 3: “The supreme excellence consists in breaking the enemy’s resistance without fighting.” The highest form of generalship, he says, is to defeat the enemy’s plans — and “the worst policy of all is to besiege walled cities.” Chapter 2 warns that “there is no instance of a country having benefited from prolonged warfare.” His whole system is preparation: know the enemy, know yourself, calculate everything in the temple before the battle is fought — so that the battle itself becomes unnecessary.</p><p>Laozi, Chapter 31: “Arms are instruments of ill omen, not the instruments of the gentleman. When one is compelled to use them, it is best to do so without relish.” Chapter 68: “A good soldier is not violent.” Chapter 30 goes further: “One who serves a ruler with the Way will never take the world by force of arms.”</p><p>Read them together and the shared doctrine is unmistakable: the most important thing about war is to be so prepared for it — so clear-eyed about its costs, so complete in your understanding — that fighting becomes unnecessary. Sun Tzu arrives there through cold pragmatism; Laozi arrives there through moral clarity. Different roads, same destination.</p><p>Scholars have noticed. François Jullien, the French sinologist, argues that classical Chinese strategy aims “to win a battle without waging war” — a structural parallel to wu wei, effortless action. And the tradition itself fused the two: in the 9th century, a general named Wang Chen wrote a commentary reading the Tao Te Ching as a strategy manual. For over a thousand years, Chinese readers have treated these as companion texts, not rivals.</p><p><strong>A civilization’s vocabulary of restraint</strong></p><p>Here’s the thought that kept me up, and the one I want to be careful with.</p><p>These two books are among the most studied texts in Chinese civilization — memorized by emperors, generals, and officials for two millennia. A culture steeped in them has, at minimum, a deep vocabulary for restraint: deterrence through preparation, victory defined as the absence of battle, conquest treated as a failure of statecraft rather than its glory.</p><p>I want to be honest about the limits of that claim, because the history is mixed. The Qin, Han, Tang, and Qing dynasties all expanded by force. Scholars like Alastair Iain Johnston have shown that the hard-edged, realpolitik strand of Chinese strategic thought was often dominant in practice. These texts supplied the language of restraint — they did not guarantee restrained behavior. A vocabulary is not a destiny.</p><p>But a vocabulary still matters. It shapes what counts as wisdom, what a ruler can justify, what the culture admires. And there is something worth sitting with in the idea that one of the world’s great civilizations spent 2,500 years teaching its leaders that the finest victory is the war that never happens — that being utterly prepared is not the prelude to fighting, but the substitute for it.</p><p><strong>The data says the substitute is winning</strong></p><p>Which brings me to the part that surprised me most, and the reason I’m writing this with hope rather than dread.</p><p>We live inside a news cycle that makes war feel constant and escalating. Zoom out to the decade scale and the picture inverts.</p><p>World War II killed an estimated 60 to 85 million people — about 3% of all humanity alive in 1940. In pre-state societies, anthropologists estimate warfare killed more than 500 out of every 100,000 people per year. Across the blood-soaked 20th century, the average was about 60 per 100,000. By the early 2000s, battlefield deaths had fallen to 0.3 per 100,000 — a decline of more than 99% from the tribal baseline, per Steven Pinker’s The Better Angels of Our Nature.</p><p>Yes, the recent years have been genuinely terrible: 2022 saw roughly 237,000 deaths in organized violence, the worst since the Rwandan genocide, driven by Ethiopia and Ukraine. But on a planet of 8 billion people, that is about 3 deaths per 100,000 — still an order of magnitude below 20th-century averages. International wars per year fell from more than six in the 1950s to less than one in the 2000s. Interstate wars now make up under 5% of all armed conflicts. And no great power has fought another directly since September 1945 — what the historian John Lewis Gaddis called the Long Peace, now in its ninth decade.</p><p>Critics dispute the statistics, and the future is not guaranteed — Pinker himself insists his data describe the past, not a promise. But the direction across a century is unmistakable, and it is counterintuitive enough to matter: the era of total war gave way to an era in which war, while still horrific where it burns, kills a shrinking fraction of us.</p><p><strong>The wars we’re actually fighting</strong></p><p>There’s a coda to all this, and it’s playing out right now.</p><p>The world is becoming multipolar — great civilizations rising in both East and West — and notice what their rivalry actually looks like. Take the last two years of US-China competition: almost none of it resembles traditional warfare. Instead the contest is being fought with leverage over strategic resources. China has rolled out export controls on rare earth minerals, gallium, germanium, and graphite — the materials the modern economy is built on. The US has answered with controls on advanced semiconductors, record energy exports, and financial pressure. Rare earth leverage versus chip and energy leverage. Each side maneuvering to be so indispensable, so well-positioned, that direct confrontation becomes unthinkable.</p><p>That is Sun Tzu’s doctrine wearing a modern suit: balk the enemy’s plans, win without fighting, make the cost of conflict so obvious that it never happens. The battlefield moved from territory to supply chains, from armies to export licenses — but the logic is 2,500 years old.</p><p>How I see this playing out: China will keep rising, guided — at least in its strategic culture — by the ancient insight that supreme preparation makes war unnecessary. And as the great powers learn to compete through influence, resources, and positioning rather than invasion, the world has a real chance to become more peaceful than it has ever been. The data already points that way. The philosophy described the destination millennia ago. We’re just now walking toward it.</p><p>Maybe Sun Tzu and Laozi were describing, 2,500 years early, the equilibrium we’re stumbling toward: a world where preparation, deterrence, and the sheer understood cost of fighting make the fighting itself increasingly rare. The best war is the one you never fight — and by the numbers, we’re fighting fewer of them than ever.</p><p>I’m Zain Aziz. These are my musings — a short daily series on technology, philosophy, fitness, and the small, real ways Muse helps me run my life a little better. This is my journaling, out in the open, in case it sparks something for you. Share feedback, especially if you disagree so we both may grow together. Follow for a new musing every morning.</p><p>Yesterday: you never wanted the software — you wanted the service.</p><p>Tomorrow: let’s see where the mind wanders to</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d95034a316bf" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[SaaS Is Dead. You Never Wanted the Software — You Wanted the Service.]]></title>
            <link>https://medium.com/data-science-collective/saas-is-dead-you-never-wanted-the-software-you-wanted-the-service-b42418f34095?source=rss-1c60abcd0d96------2</link>
            <guid isPermaLink="false">https://medium.com/p/b42418f34095</guid>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[saas]]></category>
            <dc:creator><![CDATA[Zain Aziz]]></dc:creator>
            <pubDate>Fri, 09 Oct 2026 05:04:41 GMT</pubDate>
            <atom:updated>2026-10-09T05:04:41.206Z</atom:updated>
            <content:encoded><![CDATA[<p>Software got too cheap to rent. The numbers say the build-vs-buy equation has already flipped — and most of us just haven’t updated our habits yet.</p><p>In 2011, Marc Andreessen wrote that software was eating the world. He was right, and the meal is nearly finished. But something stranger is happening now: software is being eaten by its own abundance.</p><p>Consider the arithmetic. The App Store now holds 2,172,472 apps, per Apple’s own transparency reporting. In 2025, roughly 600,000 new apps were released — up 30% in a single year. In the first half of 2026 alone, another ~560,000 arrived, putting the platform on pace to exceed one million new apps this year, according to Sensor Tower data reported by The New York Times. Meanwhile downloads grew 3% last year and 2% in the first half of this one. Supply is growing roughly ten times faster than demand.</p><p>That gap is not a blip. It is the signature of a production cost collapsing. When AI coding tools can generate a working application from a text prompt, the marginal cost of a new app approaches zero — and the store fills with the result, including, reportedly, a flood of low-effort AI-generated apps that Apple’s review process is now straining to handle.</p><h4>The subscription mountain</h4><p>Parallel to this, consider what we pay for the privilege. The global SaaS market is projected at roughly $465–489 billion for 2026 (estimates vary by forecaster). The average organization now runs 305 SaaS applications and spends about $55.7 million a year on them, per Zylo’s 2026 SaaS Management Index. BetterCloud’s 2026 survey adds the damning detail: 53% of an organization’s software licenses go an entire year without a single login.</p><p>Read that again. We are paying half a trillion dollars a year for software, and more than half of the seats are never sat in.</p><p>SaaS was built on three assumptions that held for two decades: building software is expensive, maintaining it is expensive, and the expertise to do either is scarce. You could not afford to build your own CRM, so you rented Salesforce’s. AI has now dismantled all three, and the dismantling is measurable, not theoretical.</p><h4>The evidence, in order</h4><p>First, the build cost. Anthropic’s CEO argued at Davos this year that the old logic — “amortize a piece of software across millions of users” — may simply stop being true. Andreessen Horowitz calls the result “disposable software”: applications cheap enough to build, use, and throw away. The irony is instructive — Cursor, an AI coding tool, reportedly became the fastest B2B company in history to reach $1 billion in annual recurring revenue. The best-selling SaaS product of the AI era is a tool for not buying SaaS products.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*II1Y5tZnBzZ6WDhbAMM-zw.png" /></figure><p>Second, the maintenance burden. The operational work of running software — updates, monitoring, backups — is exactly the kind of routine, checkable labor that AI agents handle well. My own automations watch each other: one job runs the daily newsletter, a second job exists solely to verify the first is still healthy. The ops team, for personal-scale software, is now a cron schedule.</p><p>Third, the expertise moat. Retool’s February 2026 survey of 817 builders found that 35% of teams had already replaced the functionality of at least one SaaS tool with something built in-house, 78% plan to build more custom tools this year, and 93% of builders are using LLMs to code or automate at work. (Retool sells internal tooling, so treat the magnitudes as directional — but the direction is unambiguous.)</p><h4>Field notes from a $0/month stack</h4><p>I can report this from the inside, because I have been running the experiment on myself for months.</p><p>My workout logger is not an app I pay $10 a month for. It is a small web app my AI assistant built in an afternoon, deployed to Vercel, preloaded with my training plan, wired to my Apple Health data so the watch confirms what happened and the logger records what was in it. My morning AI briefing is not a subscription; it is a scheduled research job with a watchdog. My birthday-party RSVP system, my book library, my lake-house listing scanner, my concert alerts — each one is a small, purpose-built tool that costs essentially nothing to run and knows exactly what I need because it was built for exactly one user: me.</p><p>None of these would survive as products. That is precisely the point. They did not need to be products. The economics of software used to require that every tool amortize its development cost across thousands of strangers. That constraint is gone, and with it goes the entire reason to rent generic software for personal needs.</p><h4>You never wanted the UI</h4><p>Here is the deeper claim, and I think it is the correct one: you never wanted the software at all. You wanted the service — the outcome.</p><p>Nobody ever wanted a workout-logging app. They wanted the knowledge of what they lifted last time and whether they are progressing. Nobody wanted a news app; they wanted to be informed in five minutes. The interface was a delivery vehicle for a result, and we tolerated the vehicle because there was no other way to get the result.</p><p>An AI assistant with context changes this. Mine knows my calendar, my health data, my files, my habits, my standing instructions — what I have started calling its developing soul, the accumulated model of me that makes it more useful every month. When the assistant itself can do the thing — check the watch, log the set, scan the listings, draft the reply — the app in the middle becomes pure overhead. Why would you open a fitness app, navigate its menus, and fight its upsells, when a sentence in chat produces a better artifact than the app ever did?</p><h4>Which categories die first</h4><p>Apple organizes the App Store into 24 app categories (plus games). The vulnerability is not evenly distributed.</p><p>First to go: single-player utilities. Productivity, Utilities, Health &amp; Fitness, Lifestyle, Education, Reference, Photo &amp; Video — categories where the value is a function applied to your own data, with no network on the other side. A to-do app, a habit tracker, a calorie counter, a PDF scanner: these are afternoon builds now, and they will be built for an audience of one, exactly to spec.</p><p>What survives: anything with a network effect or a trust moat. Social networks, marketplaces, multiplayer games — the value is the other people. Regulated finance and health — the value is compliance, custody, and liability, which no afternoon build can replicate. Professional creative tools with decades of workflow depth. Distribution, trust, and other humans: those were always the real moats. Code never was.</p><h4>The honest counter-argument</h4><p>A serious argument must survive its strongest objection, so here it is. The Jevons paradox: when a resource gets cheaper, we consume more of it, not less. Cheaper software may mean more software spending, not less — and indeed forecasters still project the SaaS market growing at double-digit rates through the decade. There is also the “6% fallacy”: coding is only about a quarter of engineering time, and R&amp;D only 10–25% of a mature SaaS company’s revenue, so AI’s cost savings touch a small slice of the P&amp;L. Sales, support, security, and compliance do not get cheaper just because code does.</p><p>Both objections are valid — and both miss the specific claim. The claim is not that the software industry shrinks. It is that the per-seat license for commodity functionality stops being the way value is captured. What replaces it has a name already: service-as-software. AI-first firms that sell the outcome, not the tool — one managed-security provider reportedly reached $40M in annual revenue in a single year on this model. The subscription does not die; it migrates from the interface to the result.</p><h4>The question flips</h4><p>For forty years, the consumer question was “which app should I subscribe to?” The question replacing it is stranger and better: “why am I renting what I could own?”</p><p>The App Store will not empty out. It will fill further — a million new apps this year, most of them forgettable. But increasingly, the software that matters in your life will not come from the store at all. It will be built for you, by something that knows you, at a cost so low that charging for it would be absurd. The store was a distribution solution to a production problem. The production problem is over.</p><p>I’m Zain Aziz. These are my musings — a short daily series on technology, philosophy, fitness, and the small, real ways Muse helps me run my life a little better. This is my journaling, out in the open, in case it sparks something for you. Share feedback, especially if you disagree so we both may grow together.</p><p>Yesterday: the AI newsletter that never misses a day.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b42418f34095" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science-collective/saas-is-dead-you-never-wanted-the-software-you-wanted-the-service-b42418f34095">SaaS Is Dead. You Never Wanted the Software — You Wanted the Service.</a> was originally published in <a href="https://medium.com/data-science-collective">Data Science Collective</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My Personal Trainer Lives in My Watch]]></title>
            <link>https://medium.com/@zanyaziz/my-personal-trainer-lives-in-my-watch-f6bce15c049a?source=rss-1c60abcd0d96------2</link>
            <guid isPermaLink="false">https://medium.com/p/f6bce15c049a</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[health]]></category>
            <category><![CDATA[coding]]></category>
            <category><![CDATA[meta]]></category>
            <dc:creator><![CDATA[Zain Aziz]]></dc:creator>
            <pubDate>Tue, 06 Oct 2026 14:55:08 GMT</pubDate>
            <atom:updated>2026-10-06T14:55:08.225Z</atom:updated>
            <content:encoded><![CDATA[<p>I kept meaning to work out more regularly. Then I gave my AI access to the one witness that never lies about it: my Apple Watch and its vitals tracker.</p><p>I have a home gym. That was supposed to solve the problem. No commute to a gym. No waiting for equipment. Just walk in, train, done.</p><p>And yet the hardest part was never the workout. It was the week quietly getting away from me — a missed session here, a “I’ll do it tomorrow” there — until I honestly couldn’t tell you whether I’d trained twice that week or not at all.</p><p>Memory is generous that way. It rounds in your favour. My watch which has a vitals tracker doesn’t. A plan small enough to actually happen. We kept the plan deliberately boring.</p><p>Monday night is tennis — that’s been my standing game for a while, and it’s the session I protect first.</p><p>Then two strength sessions a week, built around the equipment I already own: squat-and-push one day, hinge-and-pull the other. Goblet squats and dumbbell presses. Romanian deadlifts, rows, carries, planks. Roughly 45 minutes, at home, no heroics.</p><p>Each strength day has a fallback day, because life happens and a plan that breaks the first time a meeting runs late isn’t a plan — it’s a guilt generator.</p><p>Nothing about this is exotic. That was the point. I didn’t need a more impressive programme. I needed the one I already had to actually happen, every week, without me having to think about whether it had.</p><p>The check-in changed everything</p><p>Here’s the part that actually moved the needle: Muse can see my Apple Health data.</p><p>Not in a creepy, dashboard-of-my-body way. In a very simple way: did a workout happen today, or didn’t it? So we set up a rhythm. After my Wednesday and Saturday sessions, Muse checks the watch data and checks in. If a workout’s there, it logs it and we move on.</p><p>I used to type my weights into chat like some kind of telegram operator. Now there’s a Session Logger — the check-in points me to it, the day’s session is preloaded, and I tap through set by set: weights, reps, RPE, a note. No chat arithmetic, no lost numbers, no “what did I lift last time?” The watch confirms the workout happened; the logger remembers what was in it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/660/1*TM9CtRolHD4EquowlQJhtg@2x.jpeg" /></figure><p>And if 7 PM rolls around on a workout day and the watch shows nothing? I get a nudge — with the short version of that day’s session attached, so “what was I supposed to do again?” never gets to be the excuse.</p><p>Sunday works as a backup: if Saturday got missed, it checks whether anything happened, and the week gets rescued instead of written off.</p><p>It’s a small thing, but it flipped the accountability completely. I’m no longer reporting my workouts from memory, generously edited. The watch reports them. I just have to make the report a good one.</p><p>My favourite moment so far wasn’t a workout Muse planned. It was one it noticed.</p><p>I played an extra tennis session one Saturday morning — not a scheduled strength day, just a game. Muse picked it up from the watch on its own and asked the obvious, human question: does this count as today’s workout, and should we skip the evening strength reminder?</p><p>That’s the whole relationship in one message. It saw what actually happened, it didn’t blindly follow the schedule over the reality, and it left the call with me.</p><p>The reverse has happened too. A busy week where I was on my feet plenty, moving around all day — and the watch showed daily activity but no actual structured workouts. There’s a difference between a busy week and a trained week, and it’s very easy to blur the two when you’re the one grading your own homework. The data doesn’t blur them.</p><p>No lectures, either. That matters. The check-ins are short: here’s what showed up, here’s what’s next. I asked for a trainer, not a motivational speaker.</p><p>Every other fitness system I’ve tried put the burden of honesty on me. Log the workout. Track the set. Remember the week.</p><p>This one just watches the watch — and lets me spend my willpower on the actual training instead of the bookkeeping around it.</p><p>Consistency was always the programme. Now something other than my optimism is keeping score.</p><p>I’m Zain Aziz. These are my a-musings — a short daily series about the small, real ways I use Muse to run my life a little better. Yesterday: I stopped listening to my favourite podcast, and I’m more caught up than ever. Tomorrow: the AI newsletter that never misses a day.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f6bce15c049a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Stopped Listening to #1 Podcast in the world, yet I’m More Caught Up Than Ever.]]></title>
            <link>https://medium.com/@zanyaziz/stopped-listening-to-1-podcast-in-the-world-yet-im-more-caught-up-than-ever-3d4366c57531?source=rss-1c60abcd0d96------2</link>
            <guid isPermaLink="false">https://medium.com/p/3d4366c57531</guid>
            <category><![CDATA[podcast]]></category>
            <category><![CDATA[personal-story]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Zain Aziz]]></dc:creator>
            <pubDate>Sun, 04 Oct 2026 16:37:35 GMT</pubDate>
            <atom:updated>2026-10-04T17:25:03.991Z</atom:updated>
            <content:encoded><![CDATA[<p>The All-In Podcast is one of my favorites. I just don’t have 90 minutes every week for it. So I’d take shortcuts; put it on before sleeping, break it out into multiple car rides and sometimes just miss it period</p><p>There’s a specific kind of guilt that only podcast listeners know.</p><p>A new episode drops. You really do want to hear it — the market take, the AI debate, the argument that’s going to get clipped all over the internet that weekend. You save it for a drive, a workout, a walk that never quite gets long enough.</p><p>Then next week’s episode drops on top of it.</p><p>I follow the All-In Podcast on purpose. Jason Calacanis, Chamath Palihapitiya, David Sacks, and David Friedberg talking tech, markets, and AI is exactly the kind of conversation I want to stay current on. What I don’t have is a standing 90-minute block every week to sit through the whole thing, poker stories and all.</p><p>So I stopped trying to listen to every episode end-to-end.</p><p>I have Muse summarize it instead — and weirdly, I’m more caught up now than when I was “saving it for later.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/660/1*20AsYu5ayM-8u1O7wgyXAw@2x.jpeg" /></figure><h3>Saturday morning, it’s just there</h3><p>Every Saturday morning, Muse checks the All-In YouTube channel for a new full-length episode. If there’s a new one, I get a digest. If there isn’t, I hear nothing.</p><p>Silent on quiet weeks. That part matters more than it sounds. I don’t need another notification telling me nothing happened.</p><p>When there is an episode, it comes in two sections, and the split is the whole point:</p><p>1. The events they cover<br>Just the facts. What was announced, what happened in markets, what companies and stories they walked through. No spin layered in yet.</p><p>2. The All-In take<br>Their opinions — clearly labeled as opinions. Who was bullish, who pushed back, where they disagreed with each other.</p><p>That separation is why I trust the shortcut. A bad summary blends reporting and hot takes until you can’t tell which is which. This one doesn’t. I get what happened, then I get what four very opinionated people think about what happened, and I can make up my own mind in between.</p><p>Reading it takes about five minutes.</p><h3>The 90 minutes I didn’t spend</h3><p>Here’s the honest math. One episode I might skim at 1.5x speed, half-listening while doing something else, retaining maybe the big argument. Or five focused minutes with the facts and the takes separated — and I actually remember it on Monday.</p><p>Am I getting everything? Of course not. I’m missing the jokes, the tangents, the chemistry that makes the show the show. If an episode really hooks me from the summary, I go listen to that segment properly. That happens now because I have the summary, not in spite of it. It became my filter, not my replacement.</p><p>That’s changed how I think about “keeping up.” Keeping up was never about consuming everything. It was about knowing what happened, knowing what smart people are arguing about, and knowing which parts deserve my actual time.</p><p>The digest gives me the first two in five minutes. I spend the saved hour on the third only when it’s earned.</p><h3>A show I follow, on a schedule I control</h3><p>This started in September, when I asked Muse to build me a digest page starting with the September 11th episode. We settled on the facts-versus-opinion format first — I approved the structure before it ever became a weekly routine — and then it became a Saturday check that just runs.</p><p>That order matters to me: approve the format, then automate it. Not the other way around.</p><p>There’s a version of this where an AI watches everything for you and you slowly stop having your own opinions about anything. That’s not what this feels like. It feels like the newspaper version of a podcast I already chose to follow — a front page, not a substitute brain.</p><p>I still have favorite hosts. I still disagree with takes. I’m just not three episodes behind anymore, pretending I’ll catch up on a long drive someday.</p><p>Someday never came. Saturday morning does.</p><p>I’m Zain Aziz. These are my musings — a short daily series about the small, real ways I use Muse to run my life a little better. Yesterday: my morning commute became a book club. Tomorrow: the personal trainer who keeps tabs on me through my watch.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3d4366c57531" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My Morning Commute Is Now a Book Club]]></title>
            <link>https://medium.com/@zanyaziz/my-morning-commute-is-now-a-book-club-819957afc735?source=rss-1c60abcd0d96------2</link>
            <guid isPermaLink="false">https://medium.com/p/819957afc735</guid>
            <category><![CDATA[books]]></category>
            <category><![CDATA[meta]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[musings]]></category>
            <dc:creator><![CDATA[Zain Aziz]]></dc:creator>
            <pubDate>Sun, 04 Oct 2026 07:58:47 GMT</pubDate>
            <atom:updated>2026-10-04T08:10:14.866Z</atom:updated>
            <content:encoded><![CDATA[<h4>I wanted to read more despite not having time. So now a book shows up every morning at 7 — as a podcast I can listen to and a summary I can actually finish.</h4><p>I have a list problem.</p><p>Not a reading problem, exactly. A &quot;someday I&#39;ll get to the classics&quot; problem. The books everyone says changed how they think — Nietzsche, Dostoevsky, Plato — sitting in a mental pile while my commute stays… a commute.</p><p>So I asked Muse to turn that pile into a routine.</p><p>Check it out: <a href="https://zanyaziz.github.io/zains-book-library/">https://zanyaziz.github.io/zains-book-library/</a></p><p>We started with a &quot;Top 10 Books of All Time&quot; list. One book a day. Not the whole book — I&#39;m not finishing The Brothers Karamazov before work — but a real, honest version of it: a podcast episode I can listen to on the drive, plus a 20-minute read summary for when I want the details on paper.</p><p>Every morning around 7 AM, it&#39;s just… there.</p><h3>Two AI generated co-hosts personalized to my taste</h3><p>Muse writes the episode, then two voices — Maya and Dev AI generated characters— host it like a morning show. Same cast every day, which weirdly matters. It stops feeling like generated audio and starts feeling like a show I follow.</p><p>By October 3rd we&#39;d done eight:</p><ol><li>Thus Spoke Zarathustra — Nietzsche</li></ol><p>2. The Brothers Karamazov — Dostoevsky</p><p>3. The Conquest of Happiness — Bertrand Russell</p><p>4. The Republic — Plato</p><p>5. The Alchemist — Paulo Coelho</p><p>6. Beyond Good and Evil — Nietzsche again, no regrets</p><p>7. Thinking, Fast and Slow — Daniel Kahneman (I moved this one up myself)</p><p>8. Crime and Punishment — Dostoevsky</p><p>Episode 9, Sapiens, is next. Then The Art of War, the Tao Te Ching, and a block I picked myself — The Forty Rules of Love, A Gentleman in Moscow, 10% Happier (deliberately before The Power of Now), Nexus, The Anarchy, 1984. Eighteen books deep and I haven&#39;t had to think about &quot;what&#39;s next&quot; once.</p><h3>It got a website, because Apps are dead and over rated</h3><p>At some point I wanted to re-listen to an episode and had to go digging. So Muse built a library site — every episode with a play button and its summary PDF, clean and minimal. It&#39;s live at https://zanyaziz.github.io/zains-book-library/, and I liked the design enough to say so out loud, which I don&#39;t do often.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*i70hpiCJUurA4xO31UMhaw.png" /><figcaption>The Zains Morning Commute Books library — every episode with audio player and 20-min summary PDF. Listen: https://zanyaziz.github.io/zains-book-library/</figcaption></figure><p>Now the daily job updates it automatically: new episode drops, the MP3 and PDF land on the site, the page rebuilds, and Muse checks the live page actually updated before calling it done.</p><p>There&#39;s even an hourly recovery check. One morning the audio failed and only the PDF landed — Muse told me plainly that half was missing, fixed it that evening, and now a background check retries any stuck episode on its own. It only bothers me if something&#39;s been stuck for more than a day.</p><p>That&#39;s the whole deal I want from automation: handle it, verify it, and only talk to me when there&#39;s a real decision.</p><h3>What stuck with me</h3><p>I didn&#39;t find more time to read. I stopped needing to.</p><p>The commute was already happening. The books were already on a list. Muse just connected the two and then — this is the part I keep noticing — took responsibility for the boring reliability work: the queue, the uploads, the &quot;did it actually publish?&quot; check.</p><p>Eight books in about a week. Not skimmed listicles. Books I can now actually talk about.</p><p>Tomorrow morning, Sapiens will be waiting. I&#39;ll press play and drive.</p><p>I’m Zain Aziz. These are my daily a-musings — a short daily series about the small, real ways I use Muse to run my life a little better. Yesterday: I asked AI to find a birthday venue, and it built my son a website instead. Tomorrow: the AI newsletter that never misses a day.</p><p>Links: https://zanyaziz.github.io/zains-book-library/</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=819957afc735" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>