<?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 Ashok Raj on Medium]]></title>
        <description><![CDATA[Stories by Ashok Raj on Medium]]></description>
        <link>https://medium.com/@ashok-raj2520?source=rss-d4456bf817c4------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*hoGMTgrUaEOULx6blzYerA.png</url>
            <title>Stories by Ashok Raj on Medium</title>
            <link>https://medium.com/@ashok-raj2520?source=rss-d4456bf817c4------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 15:55:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ashok-raj2520/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[The Myth of “Agents”]]></title>
            <link>https://ashok-raj2520.medium.com/the-myth-of-agents-b31387625e60?source=rss-d4456bf817c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/b31387625e60</guid>
            <dc:creator><![CDATA[Ashok Raj]]></dc:creator>
            <pubDate>Thu, 10 Sep 2026 09:08:47 GMT</pubDate>
            <atom:updated>2026-09-10T09:08:47.033Z</atom:updated>
            <content:encoded><![CDATA[<p><em>There is no such thing as an AI agent. There is a model, a loop, some tools, and a costume — and the industry has been selling you the costume.</em></p><h3>The word is doing a lot of work</h3><p>Open any vendor deck this year and count the nouns. Coding agents. Review agents. Security agents. Agent swarms, agent marketplaces, agents hiring agents. The word carries a quiet promise: something with judgment, memory, skill, and accountability — a colleague, roughly, that happens to run on a GPU. That promise is the myth. Not because the systems don’t work — many work impressively — but because “agent” names a <em>packaging</em>, not a <em>thing</em>, and when you govern a packaging as if it were a person, you make exactly the mistakes the packaging was designed to make you make.</p><h3>Take one apart</h3><p>Decompose any “agent” and four parts fall out, none of which is an agent:</p><ul><li><strong>An engine</strong> — a model behind an endpoint. Pure capability, zero identity, replaced quarterly. The engine does not know it is a “security agent”; it knows the tokens in front of it.</li><li><strong>A loop</strong> — code that sends context, receives tool calls, executes some of them, and repeats until a stop condition. A while statement with a budget. Codex, Claude Code, Copilot, LangGraph are loops with opinions.</li><li><strong>Tools</strong> — functions the loop is allowed to call. The only place effects on the world actually happen.</li><li><strong>A costume</strong> — a persona prompt, a skill library, a name. “You are a senior security reviewer.” The costume is where the specialization claim lives, and it is a text file.</li></ul><p>Google’s own engineering guidance puts the split at roughly ten percent model, ninety percent harness. One team moved a coding agent from outside the top thirty to the top five on a public benchmark by changing only the harness — same engine, different loop and tools. Nobody promoted the engine. Nobody hired the agent. They rewrote the packaging.</p><h3>Why the myth persists</h3><p>Three incentives keep it alive. <strong>Anthropomorphism sells</strong>: “an agent that reviews your code” raises a budget line; “a loop that calls a model with a persona prompt” does not. <strong>Demo economics</strong>: an agent finishing a task on stage looks like competence; the harness that made it possible is invisible, so the credit lands on the noun. <strong>Accountability laundering</strong>: “the agent did it” is a sentence that assigns responsibility to nobody — convenient for vendors, dangerous for institutions, and impossible for an auditor to accept.</p><h3>What believing it costs</h3><p>The myth has four bills, and enterprises are starting to receive them.</p><p><strong>Misplaced trust.</strong> If the agent is a colleague, its report is testimony: “tests pass, implemented as specified.” If the agent is a loop with a costume, its report is <em>output</em> — and output is verified, never believed. The empirical record is unkind to the testimony view: in one study of roughly thirty-three thousand agent-authored pull requests, 28.5% never merged, and each failed check cut merge odds by about fifteen percent. Agents say “done” fluently. Fluency is not evidence.</p><p><strong>Skill sprawl as governance.</strong> If the agent has skills, you govern it by writing more skills — five-thousand-line prompts explaining how to test, how to name, how to be careful. Instructions are advisory; models follow them probabilistically; the next model release re-rolls the dice. One evaluation of repository instruction files found they raised inference cost by a fifth and improved task success not at all. The costume got heavier. The behavior didn’t change.</p><p><strong>Vendor lock disguised as capability.</strong> The durable asset in an “agent” product is never the engine (it churns) or the loop (it’s a hundred lines of ideas everyone has); it’s the accumulated <em>configuration</em> — personas, skills, rules, context. If that lives inside a vendor’s agent, the vendor owns your specialization. If it lives in your files, you do.</p><p><strong>Authority by wardrobe.</strong> The worst bill. A costume that says “you may approve” changes nothing about whether approval occurred — but a system that treats the agent as a colleague may <em>record</em> it as if it did. This is authority laundering: a weaker source wrapped in a stronger name. It is the exact failure regulators will be reading for.</p><h3>The registry that makes the myth official</h3><p>The myth’s final form is institutional: the <strong>centralized agent registry</strong> — one enterprise catalog where approved agents live, get certified, and are provisioned to teams. It sounds like governance. It is the myth with a database.</p><p><strong>It certifies the costume.</strong> A registry entry says “Security Reviewer Agent v2.3 — approved.” Approved <em>what</em>, exactly? The engine will be swapped by the vendor next quarter; the loop will be patched; the persona prompt is a paragraph. The certification attaches to a name, and the name is the only stable part — which is to say, the registry certifies the one component that does nothing.</p><p><strong>It moves trust to the wrong place.</strong> Once an agent is “registered,” its output inherits the registry’s authority: registered agent, therefore trusted output. That is authority laundering with a corporate seal — precisely the inversion regulators will unwind first. Trust cannot be granted to a producer in advance; it can only be earned by an output, after the fact, against evidence.</p><p><strong>It becomes a bottleneck by design.</strong> Every new capability, every prompt change, every skill edit now flows through a central approval queue — the classic control-plane failure, rebuilt for agents: it governs a <em>picture</em> of the fleet (names, versions, owners) while the actual behavior drifts underneath with every model release. And when the registry is down or slow, work either stops or routes around it — and routed-around governance reports green.</p><p><strong>It is a single point of drift.</strong> One catalog, many consumers, no mechanism binding a team’s actual usage to the registered definition. The registry says v2.3; the team runs whatever their harness loaded last Tuesday. Nothing checks. Nothing can.</p><p>The honest alternative keeps the useful half and drops the dangerous one. <strong>A catalog is fine; an authority is not.</strong> Keep a <em>directory</em> — for discovery, ownership, and hygiene: who published this role, where does it live, what does it declare. But move the governing act to the <em>edge</em>: each team pins the exact role, unit, and engine it uses by hash in its own repository, every generation records all three, and admission judges the <em>output</em> — never the registration. Delete the catalog and nothing stops; every pin still resolves, every receipt still names its costume. That is the test that separates a directory from a control plane: <strong>if its disappearance changes any verdict, it had authority it should never have held.</strong></p><p>The registry wanted to be the place trust is granted. Trust was never grantable. It is the place agents are <em>listed</em> — and that is all a list should ever be.</p><h3>What’s actually there — and how to name it</h3><p>Drop the noun and four honest ones appear: <strong>engines</strong> think, <strong>units</strong> run, <strong>roles</strong> know, <strong>contracts</strong> bound. A role — persona plus skills plus rules plus curated context, versioned in a repository — is the real specialization, and it is portable across every engine and every loop. A contract — exact scope, budgets, the witnesses that will judge the result — is the real control, and it works on any of them. The “security agent” becomes <em>a security role, run by whichever loop, on whichever engine, under this contract</em> — and every one of those four is a file you can diff, review, hash, and swap.</p><p>Specialization survives the death of the agent. It just stops being a product and becomes configuration.</p><h3>The governance consequence</h3><p>Once the agent is a costume, the design of trust changes shape. You stop asking <em>“can this agent be trusted?”</em> — a question about a fiction — and ask the questions that have answers: <em>Was this output permitted for this exact work? What proves it? Who accepted the remaining risk? Is the evidence still valid for these exact bytes?</em> Those are questions about records, not about characters. Agents don’t answer them. Admission does: the output of any engine, under any costume, is a proposal until independent evidence and a human signature make it authoritative.</p><p>The industry is selling workers. What you are actually buying is engines, loops, and wardrobes — and the only thing worth owning outright is the wardrobe plus the proof. Build your governance for that, and the next model release is a config change instead of a migration.</p><p><em>The agent is the costume. The specialization is the wardrobe. The proof is the only thing that was ever real.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b31387625e60" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Outcome Mode: The End of the AI Speed vs. Enterprise Control Compromise]]></title>
            <link>https://ashok-raj2520.medium.com/outcome-mode-the-end-of-the-ai-speed-vs-enterprise-control-compromise-029332eba46a?source=rss-d4456bf817c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/029332eba46a</guid>
            <dc:creator><![CDATA[Ashok Raj]]></dc:creator>
            <pubDate>Fri, 04 Sep 2026 02:52:37 GMT</pubDate>
            <atom:updated>2026-09-04T02:59:02.605Z</atom:updated>
            <content:encoded><![CDATA[<p>How Outcome Mode replaces bureaucratic ceremony with deterministic computational proof, letting your agents run fast without breaking the rules.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Bse00D4EflMvfD-k" /><figcaption>The modern agentic SDLC requires a new architecture of control.. Source: witsanu sinlapachai / Getty Images</figcaption></figure><p>Right now, engineering teams are stuck making a terrible choice: let AI coding agents run loose, or stick to heavy, outdated delivery processes from a time when writing code was actually the bottleneck. If you lean into autonomy, you end up with quick weekend proofs-of-concept that cement terrible architecture and cause regressions weeks later. But if you double down on control, the<strong> red tape just pushes developers to bypass governance entirely</strong></p><h3>The Trap of Automating the Past</h3><p>Right now, most of the industry’s approach to agentic delivery is trapped in a fundamental error: <strong>it is trying to automate the existing SDLC with workflows.</strong></p><p>Vendors are building swarms of AI agents to fill out Jira tickets, summarize pull requests, and simulate architecture review boards. They are taking a process built entirely to manage human bottlenecks and training AI to run that same bureaucracy faster. When you apply agents to legacy workflows, you don’t remove the friction — you just scale the noise. You are automating the ceremony instead of evaluating the outcome.</p><p>The Decay of the “Big Skill File”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*T2D286kX8eW7YpNocx26jA.jpeg" /></figure><p>As part of this early panic to control AI, enterprises are falling into another trap: the “big skill file.” used in agents.Teams try to govern agents by feeding them massive, monolithic prompt files detailing exactly how to use tools, write tests, or format code step-by-step.</p><p>As frontier models evolve, this rigid, instruction-heavy approach is already turning into massive technical debt. Tomorrow’s frontier models don’t need a 5,000-line prompt telling them <em>how</em> to code. They simply need a concise instruction paired with a highly accurate, valuable world model of your repository. By tying your enterprise to a massive library of brittle agent “skills,” you guarantee your delivery system might break every time OpenAI or Anthropic drops a new model update.</p><p><strong><em>Frontier models don’t need to be told how to code; they need to be told exactly what counts and then checked. Everything else in the big skill file is technical debt with a system prompt.</em></strong></p><h4><strong><em>To understand why this is dangerous, we have to look at what happened in the cockpit.</em></strong></h4><h3>From Mechanical Cables to Bounded Outcomes</h3><p>When commercial aviation transitioned to fly-by-wire in the late 1980s, the industry faced the exact same crisis engineering teams face today with AI agents. If manufacturers had simply built robotic systems to pull physical control cables faster, they would have just automated disaster. Instead, they completely rebuilt the architecture of control, shifting the pilot from a mechanical operator to a systems manager.</p><p>In the mechanical era of aviation, pilot inputs directly moved the physical flight surfaces. In the legacy SDLC, developers directly write every line of code, and if they make a mistake, it requires another human,acting as a secondary checker ,to catch it in a pull request before production breaks.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Q6woTMzcR915awA3" /><figcaption>Fly-by-wire shifted pilots from mechanical operators to systems managers.. Source: Ollie Desforges / Getty Images</figcaption></figure><p>When Airbus introduced the A320, it replaced cables with computers. Pilots no longer commanded the physical ailerons; they commanded an <em>outcome</em> (e.g., “bank 30 degrees”). The flight computer determined how to move the surfaces to achieve that outcome. Outcome Mode is doing for the agentic SDLC exactly what fly-by-wire did for the cockpit. A developer requests an outcome (“add a cache to this service”), and the AI agent determines the specific keystrokes required.</p><p>Outcome Mode refuses this compromise. It is the lightest policy-permitted way to deliver a change — one request, one plan, one candidate, one decision — with the full standard of proof intact. Not governance removed; ceremony removed.</p><h3><strong>Flight Envelope Protection for Code</strong></h3><p>The magic of fly-by-wire wasn’t just electronic cables; it was <strong>Flight Envelope Protection</strong>. The flight computer constantly calculates the physical limits of the aircraft. If a pilot commands a maneuver that would stall the plane, the system simply refuses to exceed the safety margin — no matter how hard the stick is pulled.</p><p>Outcome Mode applies this exact concept to code:</p><ul><li><strong>Establish the Contract (The Envelope):</strong> The kernel deterministically pins the baseline, the scope (what may be read and written), permitted effects, and the required witnesses (tests, structural checks).</li><li><strong>Bounded Autonomy:</strong> Any compatible agent — Copilot, Claude Code, or a local model — works in an isolated workspace. The platform does not orchestrate the agent; it bounds it.</li><li><strong>Real-Time Risk Calculation:</strong> The kernel observes the bytes, not AI vibes. If the agent attempts to touch a protected public API, rewrite a core security module, or break the contract, it hits the envelope. The kernel halts the agent’s autonomy by arithmetic, not opinion.</li></ul><h3>Escaping Automation Surprise</h3><p>As autopilots took over the mechanical flying, aviation suffered a new problem: “automation surprise,” where pilots lost track of what the computer was doing. Aviation solved this not by removing the automation, but by introducing graceful degradation. If a system encounters a scenario it cannot safely compute, it hands control back to the human.</p><p>Outcome Mode builds this directly into the SDLC. When an agent hits a risk threshold, it doesn’t fail silently or push dangerous code. It gracefully escalates into <strong>Workflow Mode</strong>. The developer is no longer writing the boilerplate; they are managing the system. The formal rail kicks in with named approvals, and every artifact and piece of evidence already produced carries forward. Promotion only ever tightens the controls based on what the bytes demand.</p><h3>How Outcome Mode Works</h3><ul><li><strong>State the Outcome:</strong> A human defines the goal (e.g., “Add a cache to this service”).</li><li><strong>Establish the Contract:</strong> The kernel deterministically pins the baseline, scope, permitted effects, and required witnesses (tests, structural checks) based on exact system rules.</li><li><strong>Bounded Autonomy:</strong> Any compatible agent — Copilot, Claude Code, or a local model — works in an isolated workspace. The platform does not orchestrate the agent; it bounds it.</li><li><strong>Real-Time Risk Calculation:</strong> The kernel observes the bytes, not AI vibes. It recalculates risk continuously based on file paths touched, public interfaces changed, and contract effects.</li><li><strong>Dynamic Promotion:</strong> If risk thresholds are crossed, the change seamlessly escalates into formal Workflow Mode. No work is lost; controls simply tighten based on what the code dictates.</li><li><strong>Bound Evidenc</strong>e: every completion claim is tied to an independent witness — the exact test, on the exact candidate — and stale or skipped evidence is refused</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*r-q7LdnddBn3Rp-t" /></figure><h3>The Enterprise-Grade Advantage</h3><p>Outcome Mode solves the AI adoption conflict structurally. Risk-proportional process stops being a corporate guideline and becomes a mathematical mechanism.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Sr-6cGedsAKc3VlahJz0aw.png" /></figure><p><em>Outcome Mode is not “vibe coding” with a great logo, nor is it an agent platform dictating how models think. The developer is the primary customer, experiencing one request, one clear plan, and one definitive decision. Success isn’t measured in governance records produced, but in reduced human review time and zero escaped defects. Outcome Mode is how an enterprise says yes to agent speed without saying maybe to proof.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=029332eba46a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Real AI Coding Crisis Is Code Nobody Can Vouch For]]></title>
            <link>https://ashok-raj2520.medium.com/ai-slop-is-a-binding-problem-not-a-quality-problem-e7d879071770?source=rss-d4456bf817c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/e7d879071770</guid>
            <dc:creator><![CDATA[Ashok Raj]]></dc:creator>
            <pubDate>Thu, 11 Jun 2026 14:55:49 GMT</pubDate>
            <atom:updated>2026-06-11T17:18:32.894Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KnpD-IZegUohitKibx4Jyw.jpeg" /></figure><p><strong>The danger is not that AI cannot write code. The danger is that it can write code faster than organizations can attach that code to design, ownership, and proof.</strong></p><p>The first mistake is calling this a quality problem.</p><p>“Quality problem” is too broad. It sends teams toward the obvious remedies: review harder, test more, buy another scanner, add another AI reviewer, write a longer instruction file. Those things may help at the margin, but they miss the deeper failure.</p><p>AI slop is not merely bad code. It is <strong>unbound code</strong>.</p><p>It is code that may compile, pass tests, satisfy the ticket, and still fail the system. It is code generated without a durable connection to the architecture it enters, the constraints it must respect, the conventions it should preserve, or the reason it exists. It looks acceptable locally and creates entropy globally.</p><p>That is the disease. The rest is symptom.</p><h3>The numbers are no longer ambiguous</h3><p>The industry now has enough evidence to stop treating this as a vibes debate.</p><p>GitClear’s 2025 analysis of 211 million changed lines found a sharp rise in copy-pasted code and a decline in moved or refactored code; the report says 2024 was the first year in its dataset where copy/pasted lines exceeded moved lines. Sonar’s 2025 analysis, drawing on Google’s DORA research and other survey data, points in the same direction: larger pull requests, longer review times, more debugging, more structural debt. A Carnegie Mellon study of Cursor adoption across hundreds of GitHub repositories found that velocity gains were large but transient, while static analysis warnings and code complexity increased persistently.</p><p>The specific numbers vary by dataset. The direction does not.</p><p>More generated code means more duplicated code, more review burden, more warnings, more complexity, and more downstream maintenance. That does not mean AI coding tools are useless. It means the bottleneck has moved.</p><p>For decades, the scarce resource in software was writing code. AI made that part cheaper. But software delivery was never only about writing code. The harder problem was always deciding what should be written, where it belongs, how it should evolve, and whether anyone can safely depend on it.</p><p>AI did not remove that problem. It exposed it.</p><h3>Correctness is not enough</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_6A3n2kigul_ojhrfOcdTA.jpeg" /></figure><p>The common defense of AI-generated code is that it works.</p><p>Often, it does. Modern coding agents can solve real issues, generate plausible patches, pass tests, and clear happy paths. SWE-bench progress shows how quickly agentic coding capability has advanced; the official leaderboard now tracks systems resolving substantial portions of real GitHub issues.</p><p>But passing tests is not the same as belonging in the system.</p><p>The distinctive failure mode of AI-generated code is not always first-order correctness. It is second-order coherence. The code can be syntactically correct, locally functional, and stylistically plausible while still violating a boundary, duplicating a concept, bypassing an invariant, or encoding an assumption no one meant to authorize.</p><p>That is why the phrase “AI slop” is useful only if we define it precisely.</p><p>Slop is not code that looks ugly.<br> Slop is code whose relationship to the system cannot be vouched for.</p><p>It is a payment flow that works by copying logic from somewhere else instead of using the shared policy engine. It is an API handler that passes tests while introducing a second interpretation of customer identity. It is a module that looks clean in isolation but quietly reaches across a boundary the architecture depends on. It is a generated abstraction that resembles the surrounding code just enough to evade review.</p><p>The patch is defensible. The accumulation is not.</p><h3>The new bottleneck is vouching</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cV7vMolV6U7mfhsOC6j4rA.jpeg" /></figure><p>This is the strategic shift the industry has not fully absorbed.</p><p>When code generation becomes cheap, code review becomes the constraint. When code review becomes overwhelmed, assurance becomes the constraint. When assurance becomes the constraint, the valuable layer is no longer the tool that writes code fastest. It is the layer that can prove the code is safe to accept.</p><p>That is why “just review it” does not scale.</p><p>A human reviewer can catch a bad line. A senior engineer can spot a suspicious abstraction. A security tool can flag a vulnerability. But AI changes the economics. It increases the volume, size, and apparent polish of code arriving at the gate. Reviewers are no longer asked to inspect a manageable stream of human-shaped changes. They are asked to evaluate a flood of locally plausible output produced at machine speed.</p><p>The review burden grows precisely because the code often looks reasonable.</p><p>Bad code that looks bad is cheap to reject. Bad code that looks good is expensive. It forces humans to reconstruct context: Why is this change here? Which boundary does it touch? Is this the approved path? What invariant is being assumed? Has this concept already been modeled elsewhere? What risk is being introduced?</p><p>That reconstruction is the work AI did not do.</p><h3>Detection is not enough</h3><p>The market’s answer to AI slop is detection.</p><p>That is understandable. If AI produces more code, buy AI tools to review it. If agents create more pull requests, add agentic reviewers. If developers miss subtle problems, ask another model to inspect the first model’s output.</p><p>Some of this is useful. But as a strategy, detection-after-generation is structurally weak.</p><p>A probabilistic reviewer judging probabilistic output is still operating in the realm of opinion. It may be better than nothing, and it may catch many defects, but it does not solve the binding problem. It still asks, after the fact, whether a piece of code seems acceptable.</p><p>The stronger question is different:</p><p>Was this code allowed to exist in this form in the first place?</p><p>That is not a vibes question. It should not depend mainly on another model’s judgment. It should be answered by constraints: architecture rules, dependency boundaries, ownership maps, API contracts, policy checks, test obligations, threat models, and traceable design decisions.</p><p>In other words, the answer to AI slop is not better taste at the end of the pipeline.</p><p>It is fewer degrees of freedom at the start.</p><h3>The markdown control plane is already failing</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wuh0ntUM1XF1v67cjpXukQ.jpeg" /></figure><p>There is a second attempted cure, and it is failing more quietly.</p><p>As teams realize agents need context, they have reached for the most available medium: prose. So repositories accumulate instruction files: AGENTS.md, CLAUDE.md, .cursorrules, skill files, custom prompts, and long lists of “do” and “do not” rules.</p><p>The instinct is right. The mechanism is weak.</p><p>AGENTS.md itself is now described as a simple open format for guiding coding agents and is used across tens of thousands of open-source projects. But the practice has already developed the predictable pathology of every unmanaged control document: it grows. Every failure adds a rule. Rules are rarely removed. Context files become sedimentary layers of old incidents, stale conventions, conflicting preferences, and one-off warnings.</p><p>Thoughtworks has already named this failure mode “agent instruction bloat”: context files accumulate, become long, conflict with each other, and bury important rules where models are less likely to attend to them. Chroma’s context-rot work found that model performance becomes less reliable as input length grows, even on controlled tasks. The “lost in the middle” research showed that models often perform worse when the relevant information is buried in the middle of long context.</p><p>The most damaging evidence comes from the AGENTS.md evaluation literature itself. An ETH Zurich study found repository-level context files produced no improvement in task success while increasing inference cost by more than 20%; it also found that agents often followed the instructions, but the instructions still did not help.</p><p>That last detail matters.</p><p>Compliance was not the problem. Governance was.</p><p>The agent read the rule. The rule was not enough.</p><h3>Suggestion is not enforcement</h3><p>A markdown file is a suggestion. A harness is a constraint.</p><p>That distinction is everything.</p><p>A suggestion says: “Please follow this architecture.”<br> A constraint says: “This change cannot merge if it violates the architecture.”</p><p>A suggestion says: “Use the shared authorization path.”<br> A constraint says: “No new authorization logic may be introduced outside the approved module.”</p><p>A suggestion says: “Keep PRs small.”<br> A constraint says: “This task may only touch these packages unless the plan is re-approved.”</p><p>A suggestion says: “Run the tests.”<br> A constraint says: “The receipt must show which tests ran, what changed, what passed, what failed, and what risk remains.”</p><p>Prose can guide. It cannot govern by itself.</p><p>This is why the next layer of value is not prompt engineering. Prompt engineering asks how to talk to the model. That matters, but it is local.</p><p>The missing discipline is harness engineering.</p><p>Harness engineering asks a larger question: <strong>What system must surround the model so that its work is scoped, constrained, checked, and evidenced before a human is asked to trust it?</strong></p><p>Martin Fowler’s discussion of harness engineering frames the harness as the environment around the model: feedforward controls that steer the agent before it acts, and feedback controls that let it self-correct after it acts. Crucially, those controls can be computational, such as tests, linters, type checkers, and structural analysis, not merely inferential model judgments.</p><p>That is the right direction.</p><p>The model should not be the whole engineer. The model should be the reasoning core inside a governed delivery loop.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PWENuT8VprTN-XlIPM2Xvg.jpeg" /></figure><h3>Architecture conformance is the missing gate</h3><p>Software architecture has had a word for this problem for years: conformance.</p><p>Architecture conformance checking asks whether the implementation still matches the intended architecture. This is not a new idea. Research and tooling have long explored ways to document architectural rules and automatically check whether code violates them.</p><p>What is new is the urgency.</p><p>When humans wrote most code manually, conformance failures accumulated at human speed. With agents, they accumulate at generation speed. The problem has not changed in kind; it has changed in rate.</p><p>And yet the LLM era has mostly skipped this layer. A 2025 systematic review of LLMs in software architecture found that conformance checking remains underexplored compared with other architecture-related LLM applications.</p><p>That gap is the opportunity.</p><p>The right question is not: “Can the model produce code?”<br> It can.</p><p>The right question is: “Can the delivery system prove that the produced code conforms to the design we approved?”</p><p>That proof cannot live inside the model alone. In safety-critical domains, this is obvious. A MISRA C++ compliance study found that popular LLMs did not reliably generate MISRA-compliant code even when explicitly prompted, though some could correct specific violations when instructed.</p><p>The lesson generalizes.</p><p>Models can assist. They cannot be the authority of record for conformance. Enforcement has to live outside the model.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZCzP5jfHepmfrjzZS_RaaQ.jpeg" /></figure><h3>From context files to Context Fabric</h3><p>The mistake in today’s agent workflows is treating context as documentation.</p><p>Context is not documentation. Context is control.</p><p>What the agent sees, what it may touch, what it must preserve, what it must prove, and when it must stop are not onboarding notes. They are operational boundaries.</p><p>A better system does not dump the whole codebase, the whole architecture document, and a thousand lines of instructions into a context window. It assembles the right context for the task and phase. It exposes the relevant symbols, dependencies, policies, tests, and decisions. It withholds what is irrelevant. It narrows the action space. It checks the result mechanically. It records what happened.</p><p>That is the layer I call <strong>Context Fabric</strong>.</p><p>Context Fabric is not a smarter prompt. It is a binding layer.</p><p>It binds agent work to five things:</p><p><strong>Intent:</strong> Which approved task, requirement, defect, or design decision authorizes this change?</p><p><strong>Context:</strong> Which files, symbols, APIs, constraints, tests, and architectural rules were provided to the agent?</p><p><strong>Permission:</strong> What was the agent allowed to read, edit, execute, create, or ignore?</p><p><strong>Conformance:</strong> Which deterministic checks prove the change respects the relevant boundaries?</p><p><strong>Receipt:</strong> What evidence exists that the change was made, tested, reviewed, and risk-assessed?</p><p>The receipt is the critical artifact.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rYBqPC9Luy6QfWTbnOsrtg.jpeg" /></figure><p>A pull request says, “Here is the diff.”<br> A receipt says, “Here is why this diff exists, what it was allowed to affect, what evidence it produced, and which checks make it safe to consider.”</p><p>That is the difference between trusting an agent and auditing one.</p><h3>Better models will not remove the need for binding</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9K8qtXESYfsQISBiEG2RaA.png" /></figure><p>The serious objection is obvious: won’t better models fix this?</p><p>They will fix some of it. They will reduce syntax errors, improve test repair, generate cleaner patches, and follow instructions more reliably. They will become better local programmers.</p><p>But the binding problem is not a local programming problem.</p><p>Architecture is a system property. Accountability is a system property. Auditability is a system property. Compliance is a system property. A model can participate in those systems, but it cannot replace them by becoming more fluent.</p><p>DORA’s 2025 research is useful here because it frames AI as an amplifier: it magnifies an organization’s existing strengths and weaknesses rather than magically repairing the underlying system. Strong engineering organizations can absorb AI because they already have clear architecture, fast feedback, reliable tests, healthy platforms, and disciplined review. Weak systems do not become strong because a model writes code faster. They become weak faster.</p><p>That is why waiting for better models is the wrong strategy.</p><p>Better models reduce local mistakes. They do not eliminate the need for external constraints. In fact, the better and faster the models become, the more important the binding layer becomes.</p><p>Acceleration without binding is just entropy with a better interface.</p><h3>Where the value goes</h3><p>When a layer commoditizes, value moves to the adjacent constraint.</p><p>Code generation is moving toward abundance. That does not make software engineering less valuable. It changes where the value is concentrated.</p><p>The valuable layer is now assurance.</p><p>Not assurance as a checkbox. Assurance as an operating system for software change: scoped context, enforceable architecture, deterministic gates, policy-aware generation, continuous conformance, and receipts that let an organization defend what it shipped.</p><p>This will matter first in regulated industries, where the need to prove conformance is explicit. Banks, healthcare companies, aerospace firms, defense contractors, insurers, and critical infrastructure providers cannot simply say, “The agent wrote it and the tests passed.”</p><p>But it will not stay there.</p><p>Any company with a large, long-lived codebase will feel the same pressure. The two-thousand-engineer software company may not answer to the same regulator, but it answers to another unforgiving authority: time. Engineers leave. Systems age. Documentation rots. Business rules migrate from architecture diagrams into folklore. AI-generated code accelerates the gap between what the system does and what the organization can explain.</p><p>That gap is the real technical debt.</p><h3>The next discipline</h3><p>The last decade of software tooling optimized the act of writing.</p><p>The next decade will optimize the act of vouching.</p><p>We will still need models. We will still need agents. We will still need tests, scanners, reviewers, and documentation. But those pieces have to be assembled into something stronger than a pile of suggestions.</p><p>The future is not “AI writes code and humans review it.”</p><p>That is too slow, too vague, and too dependent on human attention at the exact point where human attention is most overloaded.</p><p>The future is: AI proposes changes inside a governed harness. The harness scopes the task, supplies the context, restricts the action space, checks conformance, gathers evidence, and escalates judgment only where judgment is truly needed.</p><p>That is the inversion.</p><p>Do not ask humans to out-review machine-speed generation.<br> Do not ask models to self-police by reading longer markdown files.<br> Do not confuse passing tests with belonging in the system.</p><p>Bind the work before it exists.<br> Verify it before it merges.<br> Record why it was allowed.</p><p>AI slop is not a quality problem.</p><p>It is what software becomes when generation is detached from intent, architecture, and accountability.</p><p>The cure is not better prose.<br> The cure is not more review.<br> The cure is binding.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*auGayJeGG79GGtodrmayJw.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e7d879071770" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Microbes and the consciousness]]></title>
            <link>https://ashok-raj2520.medium.com/microbes-and-the-consciousness-549cfb046f2f?source=rss-d4456bf817c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/549cfb046f2f</guid>
            <dc:creator><![CDATA[Ashok Raj]]></dc:creator>
            <pubDate>Wed, 07 Dec 2022 13:26:58 GMT</pubDate>
            <atom:updated>2022-12-07T13:32:01.288Z</atom:updated>
            <content:encoded><![CDATA[<p>Viruses are one of the most important forces of the nature .Viruses and microbes are the reasons for life and the complexity that we see in biological systems . Viruses acts as stabilising force in the nature .</p><p>Microbes uses different strategies for infection .One of the ways by which viruses facilitate its propagation is by hijaking the host’s control and command centre (brain )</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*jNlYsgeE0SOmxPPP.jpg" /></figure><p><strong>Toxoplasma parasite</strong></p><p>This is a usual parasite found in cats ,rats and other animals .To complete its lifecycle it has to enter the cat intestine .Parasite uses a clever trick for that .Parasite enters the mouse brain cell and instructs the cell to send neurotransmitter messages to other neurons . This makes the mouse fearless .The fearless zombie mouse will start exploring the world and will eventually eaten by the cats. This completes the life cycle of hte parasite .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/966/0*6ufa3sG3lTx-clYS.jpg" /><figcaption>Was Nepolen infected with Toxoplasma parasite?</figcaption></figure><p><strong>Rabies</strong></p><p>Most the stories like Dracula /Zombies/Vampires are human beings attempt to understand rabies .It was unkown till late 1800s that a virus is causing the rabies .There are many myths about rabies around the world .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*hVDAZStyqsL4oLqX" /></figure><p>Rabies is a deadly disease which affects the brain and turn people to real zombies .It causes a set of phobias like fear of water,sound ,moving things etc.</p><p>Rabies changes the behaviour and introduces rapid eye movement ,high hert beats ,palpitations .Once the virus reaches the brain death occurs in 2–3 days .</p><p>( There are 14 documented cases in which the patient was able to survive rabies .Usually this is done by medically induced coma and anti-virals .99.99% cases it will not help )</p><p>In popular myth Dracula ,the vampire bats represents Dracula .Bats are the natural reservoir of rabies .Usually bats are immune to the rabies and virus gets activated when the bat is sick or immunity of the bat is low .In the Dracula story a person is recruited to his gang when he bites them using his canine teeth (Like a dog/bat bite).The person becomes zombie after 30–40 days .Rabies also takes 30–40 days to reach the brain turn people to zombies .Once they turn to zombies they start biting others gets unnatural powers .A rabies infected person is usually chained to bed and it takes 4–5 people to control the patient .</p><h4><strong>List goes on and on.</strong></h4><p>The list of microbes that alter the behaviour of animals and plants is huge .This includes flu viruses ,covid-19 and many viruses which can cause long term damage in brain .Certain dementia ,alzheimer and cancers are due to virus infection .</p><h4>Does virus/microbes take away our free will ?</h4><p>If microbes can control our brain and change our personality ,then do we have free will or is it just an illusion created by our brain ?</p><p>Famous Physicist Niels Bohr once said “<strong>Everything we call real is made of things that cannot be regarded as real</strong>.” Similar to that any decision that we think it is a free choice can be traced back to automatic and deterministic electro-chemical reactions in brain .Sam Haris in his book called “Free will” explores this subject in great detail .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*kkzicVD2Kjzl-1qz.jpg" /></figure><h3><strong>Our viral origins.</strong></h3><p>5 to 10% of our DNA codes were acquired from viruses .This is an on going process and like computer viruses they also can add its code to our own cell DNA .Many times these are harmless and once the host dies ,They detach from the host’s DNA and again becomes virus .Viruses can exchange genetic materials with other types of viruses and form ne ones pretty quickly .Virus follow one rule is that is there are no rules.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/850/0*HSSIDe9QUojOwWp4.jpg" /></figure><h3><strong>Mitocondria</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/900/0*IMCHJ1y_To1zkJct.png" /></figure><p>Certain viruses and microbes co-evolved with cell and became part of the cell .One example of such a microbe is mitochondria .This is called the power house of the cell .Mitochondria has its own DNA .Usually in sexual reproduction this DNA is passed from mother to child .Mitochondria removed its viral replication process and that way it got place to stay and a way to go to next generation and in-turn cells got a tool to produce energy efficiently .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/720/0*58cDoL-OEn8gNwqJ" /><figcaption>Usually Mitochondria DNA is passed from mother to the child .A child get more genetic material from mother than from the father</figcaption></figure><h3><strong>Learning ability and viruses</strong></h3><p>Neurons shares lot of features with viruses. The plasticity of the neurones and the way it changes the wirings/connections with other neurons gives the learning ability to the animal .</p><p>The changes in neural wiring is initialed by a gene called Arc gene . Arc genes might be responsbile for creating higher intelligence that results in consciousThis gene works almost like a virus .They go and “infect” a nerve cell and delivered cargo that is required for new connections .The “cargo” delivered by the arc gene contains detailed plan ( chemical molecues ) to cut/make connections .If we loose this gene then our learning ability will be gone for ever . The speculation is that this gene is a virus which became part of our DNA . In a way we can say that our brain exist because of some pre-historic viruses .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/769/0*9zejjsmdtWAn6YVO.jpg" /><figcaption><strong>Comparison between virus and arc gene “infection”</strong></figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=549cfb046f2f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Salt & Pepper]]></title>
            <link>https://ashok-raj2520.medium.com/salt-pepper-6f0fd0856cb9?source=rss-d4456bf817c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/6f0fd0856cb9</guid>
            <dc:creator><![CDATA[Ashok Raj]]></dc:creator>
            <pubDate>Tue, 04 Oct 2022 09:12:17 GMT</pubDate>
            <atom:updated>2023-10-20T08:10:19.185Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/521/1*AudCKF6MDwuLURV7Zn8cdQ.png" /></figure><h3><strong>Salt</strong></h3><p>Salt was once most valuable mineral that humans extracted . In ancient greek salt used to be currency and soldiers used to get paid by salt .The word salary originated from the latin words for salt and soldier (Salt paid to the soldier = salary) .</p><h4>Why is salt so precious ?</h4><p>Animals love salt .Salt is one the most essential component in maintaining pH value and mineral salt contains traces of calcium ,magnesium and various minerals .Readily consumable salt used to be in very short supply and we are hard wired to have liking for salt . Every FMCG company uses salt to trick our brains to spend more on junk foods .</p><h4>Salt and politics</h4><p>Salt taxes were introduced by governments all over the world and this lead to many protests and political changes across the world .One of the reason behind the french revolution is attributed to the raise in slat tax .In India also Gandhi started salt satyagraha to protest against salt taxes .</p><h4>Salt as a preservative</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/560/0*8cAILk2xhpQdejoK" /></figure><p>Salt can preserve things for long .Chlorine in the salt ionises water molecules in microbes and de-stabilises the cell chemistry .</p><p>Salt helped human beings to preserve meat and fruits .This in-turn helped in long expeditions ,big wars and exploration .</p><p>Many roads were created in Europe to transport salt and salt roads created the first network of roads in Europe .</p><h3>Spices as preservatives</h3><p>Like salt ,pepper(and spices) also has some very unique capabilities .In India some of the spiciest food comes from the hottest/humid regions like Rajastan,Andhra,TN etc .</p><p>Microbes thrive in hot and humid climates .Adding spices makes it tastier and keep it fresh for few more hours .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*PmmAbWddvvEDOxBM.png" /></figure><blockquote>Karam is the old tamil word for spicy flavour .The word curry originated from it.</blockquote><h4>Evolution of spices</h4><p>Spices are very unique to tropical climates .Some of the spices requires unique climatic conditions and it is not easy to cultivate in non-tropical climate .</p><p>Chemical compounds that form spices are plant’s evolutionary tricks to keep away mammals from eating them .Usually the seeds gets digested in animal stomach and does not get dispersed .Plants evolved to create certain chemical compounds that can irritate the mucus membrane in mouth and stomach of animals .Birds doesn&#39;t have mucus membrane and can consume large amounts of chilly pepper . Additionally these compounds protects plants from insects ,fungus and other microbes .Some of the compounds are known to have strong antibiotic effects .</p><p>In India western ghats (Konkan to Kerala) is home to some the unique spices in the world .Bio-diversity in these rain forests forced the plants to evolved to various spices with strong flavours to attract birds and scare away insects .</p><h4>Black gold and shipping</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ONsi4Mb568prmtZM.png" /></figure><p>Pepper used to reach Europe from china and silk route during BC .Jewish and arabs with the help of trade winds were able to reach the Malabar cost and procure spices from south India and far east .</p><p>Arbs and Jews had the monopoly of entire pepper trade to Europe and the arbs controlled the supply chain from India .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/735/0*4HFOHguWUiyN2HSr" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*D_3cYOARFczraARr.jpeg" /><figcaption>Trade winds gave arabs easy access to the west cost of India</figcaption></figure><p><strong>Muziris</strong></p><p>Hisotry of pepper trade is incomplete without the history of Muziris .Muziri was the one of the world’s busiest port during the last century of BC and initial few centuries of AD .The location of this port is not certain .There are detailed descriptions about the port in Greek,Jewish and Arab history and folk songs .</p><p>Tamil classical works during sangham period talks about the trading vessels visiting muziris and trading pepper and gold .</p><p>Ilango Adikal the Jain poet who wrote chillpathikaram has written about the port in his classical works .He was the prince of Kodungaloor and it is speculated that the muziris is kodungaloor .</p><p>Greek produced the map of the world in BC and it has muziris plotted in south of greek</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/457/0*HLft6sbAO7rD-Ejj.jpg" /></figure><p>Arabs controlled the supply chain and Jewish traders with vast network of trade routes controlled the supply from Arabia to Europe .Rise of ottoman empire and the capture of Constantinople strained the relationship between islamic world and Europe .This forced the Europeans to find an alternative supply route for spices .</p><p>Arabs with its formidable naval power controlled the arabian sea and trade winds helped the arabs to sail faster and safer way to India .</p><p>Portuguese were the first to find the route to Malabar cost through cape of good hope in south africa . Legend is that an arab ship due to rough weather lost its way and crashed near Cape of good hope and the Portuguese caught the arabs and tortured them to find the exact route to muziris .</p><h4>Opium + Salt + Pepper</h4><p>Salt and pepper helped the Europeans to stock meat and preserve fish for even bigger expeditions .</p><p>Opium was part of spice trade .Many anthropologists are of the opinion that opium usage enabled sailors to take more risks and sail through through some of the roughest oceans around Africa .</p><blockquote>Opium wars was one of the the most ridicules war British ever initiated . Like today ,there was huge balance of payment issue with British and Chinese .British forced china to accept opium as payment for the goods and china rejected it and that resulted in war .How ridicules is that. ? ..Drugs as payment??? .</blockquote><p><strong>Change in diet and spice trade</strong></p><p>Spice trade driven expeditions introduced variety of crops to different parts of the world .Some of the crops introduced to India by Portuguese are</p><ul><li>Cashew Nuts</li><li>Pineapple</li><li>Tapioca</li><li>Green chiles (chiles that we have with lemon for good luck got introduced in 16th century 😳)</li><li>Citrus</li><li>Rubber</li><li>Papaya</li><li>Corn</li><li>Ground nuts</li><li>Tomato</li><li>Maize</li><li>Kidney Bean</li><li>Guava</li><li>Tobacco</li><li>Tea</li></ul><p>Most of these have become the part of regular Indian diet and many have become staple food for millions .Tapioca saved millions from starvation in India and Africa .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ukGaN519enFP0caO.jpg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6f0fd0856cb9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data collection ,COVID and the Butterfly effect]]></title>
            <link>https://ashok-raj2520.medium.com/data-collection-covid-and-the-butterfly-effect-2765319e5301?source=rss-d4456bf817c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/2765319e5301</guid>
            <dc:creator><![CDATA[Ashok Raj]]></dc:creator>
            <pubDate>Sun, 27 Feb 2022 10:00:24 GMT</pubDate>
            <atom:updated>2022-02-27T10:02:57.311Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>Data collection ,COVID and the Butterfly effect</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/676/0*jixkQ_n8W9ZdXqRg.gif" /><figcaption>(Credit Google Search)</figcaption></figure><p>After the Chinese communist revolution of 1948 ,Mao Zendong introduced series of measures to improve the economy of china .China was an agrarian economy and most of the people were extremely poor and malnourished .</p><p>To improve the agriculture output Mao introduced a series of steps .</p><ul><li>collective farming ( people cease to own property and everything belongs to state)</li><li>Local communist comrades with the help of officials plan everything related to the farming .These guys fresh from</li><li>Killed all the pests ( Four Pests campaign :- Killing sparrows ,mosquitoes ,rodents and flies ).Killing sparrows lead to a large increase in locust and other insects .</li></ul><h3><strong>Data collection</strong></h3><p>The changes brought in by the new communist government disrupted the agriculture sector and Mao thought with all his crazy ideas agriculture production will double .</p><p>Data collection was done by the local communist party workers and they exaggerated the figures and send to the next level in party .Every layer the party officials exaggerated figures by putting zero here and there . The numbers were consolidated and send the politburo .Figures were way more than Mao ever thought .There was huge difference between the real figures and the figures Mao got .Mao exported the “excess” and this resulted in the recorded history’s biggest famine .Famine killed more than 20 -50 million people .</p><p>The exaggerating the data to please the leaders costed dear for china and its citizens .</p><p><strong>Famine changed the dietary habit of Chinese</strong></p><p>Famine resulted in huge changes in Chinese food habits .Lack of rice and loss of live stocks lead the poor Chinese to hunt and eat wild animals .Wet markets become very common and started selling wild animals .</p><p>Famine lasted for more than 10–15 years .Wet animal market became the part of Chinese culture and continued even after the famine years .</p><p><strong>Covid and wet market</strong></p><p>It is clear that the covid originated from some wild animals and jumped to humans .</p><h3><strong>Butterfly effect</strong></h3><blockquote><strong>Wrong data collection → Wrong decision to sell the grains → Change in dietary habit → SARS/Covid etc.</strong></blockquote><p><em>NB: The pest eradication stupidity created huge ecological imbalance.This might have put lot of evolutionary pressure on many viruses to migrate to other hosts .</em></p><p>Disclaimer: “<strong><em>Raise of covid is a much more complex event and the article is not meant to give any scientific explanation to emergence of zoonotic viruses like covid .</em></strong>”</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2765319e5301" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Crypto and our society]]></title>
            <link>https://medium.com/coinmonks/crypto-and-our-society-c958467805a3?source=rss-d4456bf817c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/c958467805a3</guid>
            <category><![CDATA[society]]></category>
            <category><![CDATA[hype]]></category>
            <category><![CDATA[crypto]]></category>
            <dc:creator><![CDATA[Ashok Raj]]></dc:creator>
            <pubDate>Fri, 25 Feb 2022 01:44:20 GMT</pubDate>
            <atom:updated>2022-03-02T14:11:45.239Z</atom:updated>
            <content:encoded><![CDATA[<h3>Cryptos and our society</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*a1MxACOKD6DZqwYR.jpg" /><figcaption>Credits (Google search)</figcaption></figure><p>Crypto and related techs are promising huge changes in economy/society and political structures .</p><p>Central idea is to remove the inter mediatory third parties with an immutable chain of digital signatures combined the “hard” math problem solving (Proof of work) .Block chain effectively solves the<strong> <em>double spend</em></strong><em> </em>issue common with digital assets .</p><p>Blockchain has lot of high value use cases and has the potential to bring in fundamental changes in many industries .</p><p>This article is about crypto as replacement to traditional money and assets .</p><h3>Money</h3><p>Money is said to be one of the biggest inventions of human kind after fire .It has facilitated large scale co-operation and economic prosperity .The paper currency was first introduced in china around 1100 AD by Song Dynasty .Interestingly it vanished from china after end of mongol rule (Around 1300 AD)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/220/0*TTKadnxIslsYyb2T.jpg" /></figure><p>Marco Polo visited china around 1300 AD took the paper currency idea to europe .He landed up in jail after the return and he narrated the paper currency story to the jail officials.The officials and intellectuals during that time found it hard to accept the fact that a paper can be used as currency . It took almost 800 years after Marco Polo’s death for US to transform from gold/silver and array of bank promissory notes to single unified dollar as we see today .</p><p>Today’s paper money is a shared myth that exist in its users mind .It can be a worth less piece of paper in no time .One of the best example is how India gov made 1000/500 Rs notes worthless in one single announcement which resulted in endless sufferings (even economic slowdown) .</p><p>Money can be defined as am interim storage of value for future use.</p><h3>Vision</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/850/0*w8AhU-M5NaiZA_Px.jpg" /></figure><p>Crypto evangelists and many social media influencers claim that the crypto currencies can</p><ul><li>Replace the traditional currency .</li><li>End of centralised institutions like banks.</li><li>End of stock exchanges ( Shares can be issued as an initial coin offer in crypto world)</li><li>More power /privacy/anonymity to individuals .</li></ul><p>All of these are extremely idealistic and virtuous.The question is, will this revolution turn the tide against the traditional institutions and the nation-state concept .</p><h3>Crypto as a replacement to traditional currency</h3><h4><strong>Governments and crypto.</strong></h4><p>Governments and policy makers were slow to react and most of them were unaware of the bitcoin and its growth.The technology remained in hands of few elite programmers and university students .</p><blockquote>“Crypto is Libertarian ,AI is communist “ Peter Theil</blockquote><h4>Authoritarian governments</h4><p>China being a communist dictatorship has banned all types of crypto currencies .There is no doubt that the authoritarian centralised states hates bitcoin/cryptos . [ They all love AI/Ml and use it “<em>effectively</em>” to monitor its own citizens ,silence critics ..etc ]</p><h4>Democracies</h4><p>Most of the democratic governments allows crypto to function .The policy makers are dealing with a very something which they have never encountered earlier .There is a lack of clarity and many agencies are making contradictory statements .</p><p>Government has full control over the money in all the countries .Monetary polices are used by all governments to control the inflation ,trade deficit etc .This gave raise to an inflation based economy.An inflation based economy is required to sustain the credit system that modern economies follow.</p><p>If a farmer takes 1000 Rs loan at 10% rate then at the end of the year he may have to pay 1100 Rs back to the back .He may have to increase the price of his produce to meet the bank’s rate or he may have to invest in technology to minimise the production cost.</p><p>This brings up a bigger question about inflation based economy model .Will improvements in technology make things cheaper and he may never be able to payback his loan? .Price drop may not increase the demand for many of these items [ In this case elasticity of demand is very low].Future every nations may have to look into the deflation based economic model and adopt the system towards it .</p><p>A transition from a traditional currency to a purely digital broader less crypto currency will be highly chaotic .</p><h3>Crypto asset</h3><p>Crypto is considered to be an asset .Most of the major currencies are depreciating against crypto and that makes it very attractive for hedge fund companies to move the money from USD/Euro to cryptos.</p><p>Liquidity is one of the major concern for any assets .Crypto’s liquidity depends on how fast/easily one can convert the asset to USD or other digital assets .Centralised crypto exchanges requires KYC process and acts like banks .The anonymity promised by crypto is lost in this process .</p><p>Many assets provide constant cash flow to the investor .Crypto as asset will not be able to generate any income to the investor .Appreciation of cryptos make it attractive and cash flow may not be an important factor .To keep the momentum the crypto has keep going up else people find something more attractive and will move the money out of crypto world .Bit coin’s supply is capped at 21 Millon coins .This might help in sustaining the appreciation rate of bitcoin</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mdVLT-G0JWczDGKxfjkzdw.png" /><figcaption>Image Credits businesstoday</figcaption></figure><p>Monetary polices adopted by central banks can create liquidity crunches .This can eventually result outflow of funds from crypto world to traditional currencies . This can result in massive loss for investors .</p><p>Extreme fear rules the crypto world .There is always risk of massive loss .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8my3aHbHdR2AEO2zS5EpNQ.png" /><figcaption>Credits(<a href="https://alternative.me/">https://alternative.me/</a>)</figcaption></figure><h3><strong>Black swan events and other factors</strong></h3><p>There can be some unforeseen events which can dilute the power of nation states and bring in radical changes to society .Following events can create unpredictable changes</p><ul><li><strong>Hyperinflation: </strong>USD loosing value and stability</li><li><strong>Geo-political issues and re-alignments: </strong>Post pandemic world could be very different from the one we have today .There can be three set of nations .First one authoritative countries like China,Russia,North Korea etc .Many countries which has trade links China might join this group .Second group could be the set of liberal nations which has high degree of respect for individual rights like USA and EU .Third could be the BRICS nations(sans Russia and China) .They will collaborate more within their respective groups and crypto might bring them together in their groups.</li><li><strong>Clash of generations: </strong>Most of the institutions were made by baby boomer generation and maintained by Gen X and Gen Z .The Millennials were born to a world totally different from the previous generations . The entropy has slowly destroyed these institutions and vested interests are using it for their own gains . Many are beyond repair and may require complete rebuild with fresh outlook .Millennials might find these values and institutions incompatible with their values .<br><strong>“<em>Every generation imagines itself to be more intelligent than the one that went before it, and wiser than the one that comes after it</em>.” : George Orwell</strong></li><li><strong>Internet </strong>is not a black swan<strong> </strong>but it is<strong> </strong>one of the most disruptive forces ever created by man .Most of the old ideologies (like religion) comes to internet to die .In early 1900s people started finding ways to harness the power of electricity .Dynamo and electric motor were invented and made available for public use .Manufacturing companies of those days and their assembly lines were designed to use steam power .The adoption of electricity required radical changes and disrupted the then industrial world.Similarly internet and various technologies around it has not been adopted fully and the adoption of it will require complete overhaul of laws and institutions.DiFi,DAOs ,tokenized web needs complete restructuring or removal of traditional institutions and laws</li><li><strong>Death of democracies : </strong>Democracies are slow and costly to maintain .Democracy works on checks and balances of multiple institutions .Over the period time the elected executives started getting more powers and this lead to the erosion of values of other institutions . Raise of social media gave raise to populist leaders .Most of these leaders ride on right wing nationalism and religious extremism .To stay in power they need to feed more nationalistic and religious propaganda to the masses .This will eventually lead to facist dictatorship and collapse of society .Collapse of democracies might lead to a liberal singleton government which includes countries with shared values with unified crypto currency</li></ul><p><strong>Decentralisation and human society</strong></p><p>Human beings evolved from hunter gather society to larger tribes, agrarian societies ,small nations and to the big nations that we see today .Collaborating in large groups is unique to humans (Ants and bees do co-operate in big numbers ,But they will not over throw the queen and establish a political order) . Yuval Harari has described these characteristics of human beings in detail in all his three books .</p><p>Is decentralised approach incompatible with human society? or Do we need a new approach for co-operating in large groups through internet ? .This is an open question and there are many scientific (and not so scientific) theories around the formation of proteins and life from simple compounds to a multi cellular organised life forms with large societies .</p><p><strong>Hype cycles and final adoption of blockchain.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FT9ZSAYqxMxvnk95YHvmSw.jpeg" /></figure><p>There is no doubt that the block chain can ultimately transform many process that exist today .Looking back at the history we will find lot of such innovations which created unrealistic expectations which lead to a crash and then the eventual adoption of the technology</p><ul><li><strong>1920s</strong> combustion engine invention lead to massive flow of funds to every company that had “motors” in its name.The bubble got burst and a handful of them survived .Combustion engine is an extremely useful innovation ( So good that the design has not changed much for almost 100 years) ,But the huge expectation lead to creation of many companies and created a bubble .</li><li><strong>1960s </strong>innovation<strong> </strong>in integrated circuits lead to another bubble .Any company with “tronics” in the end got lot of investments and finally that bubble also burst .After the burst Intel,Texas instruments and created a digital and computing revolution</li><li><strong>2000s </strong>.dot com crash .Crash lead to adoption of internet in a better way (Google,Amazon,PayPal etc) .</li><li><strong>2022 and beyond .</strong> There are lot of hype around many useful techs like AI/ML ,Mobile apps ,Crypto .Many mobile apps are existing for more than 8–10 years and they are yet to break even .Future income expectation lead to a huge inflow of funds to these companies and there by created multiple asset bubbles .Federal reserves highly controversial monetary polices added fuel to fire .Next crash will help in bringing value to these technologies .</li></ul><h3><strong>Animal Farm and Crypto revolution</strong></h3><p>A discussion on change in society is never complete without discussing Orwell and his works.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/893/0*z0eSn30JX7gijA_C.jpg" /></figure><p><strong>Animal farm</strong> is a satirical novel by George Orwell based on Russian revolution of 1920s .</p><p>Animal farm starts with animals in a farm planning to overthrow its cruel masters and establish a rule based on equality .</p><p>Pigs were the smartest and they constantly talked about the need for change and revolution (Current world crypto evangelists 😁) .</p><blockquote><em>Beasts of England, Beasts of Ireland,<br>Beasts of every land and clime,<br>Hearken to my joyful tidings<br>Of the Golden future time.Soon or late the day is coming,</em></blockquote><blockquote><em>Tyrant Man shall be o’erthrown,<br>And the fruitful fields of England</em></blockquote><blockquote><em>……<br>All must toil for freedom’s sake.Beasts of England, Beasts of Ireland,<br>Beasts of every land and clime,</em></blockquote><blockquote><em>Hearken well, and spread my tidings<br>Of the Golden future time</em></blockquote><p>Revolution started with a simple slogan “<strong>All animals are equal </strong>“ (<em> I don’t think there is anyone who has not heard this quote </em>). They overthrew the human masters in an over night revolution and establish their own rule.</p><p>Pigs being smart was assigned with the complicated management of farm with the help of dogs .Suddenly they realise that it is difficult to manage the farm without laws .Pigs starts creating the laws and all animals were made to believe that the laws are created to protect them and their rights .Napoleon is the leader among pigs and he establishes his own dictatorship .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bKpNVAlHZRHUgFd0.jpg" /></figure><p>Pigs slowly become more powerful and many classes emerge in the animal society .The new social order started looking like the order that they had beofore the revolution .Pigs replaced humans and dogs become slaves to pigs .All the other animals either accepted the fate or silenced by pig-dog elite class .</p><p>Pigs amend the constitution and adds a new line into it .</p><p>“<strong>All animals are equal But some are more than equals </strong>“</p><p>The society started on a very liberal ideas and equality ended in more cruel and unequal .This repeats over and over again in most of the revolutions (recent example is arab spring . Pharaohs were replaced with another set of pharaohs )</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/570/0*hpHNHVCcZkIg8qyT.png" /></figure><h3>Conclusion</h3><p>It is difficult to predict the future of crypto and many of my views are contradictory to each other .</p><p>On one hand there is a need for immutable ledger which has potential to transform various old practices and institutions and on the other hand the hype and evangelism around it .</p><p>Human brain has a unique capability to entertain two opposite views in same time .I think we need to live with this ambiguity for some more time .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/850/0*JymXx4NElEUIPI6N.jpg" /></figure><blockquote><strong>Thanks for reading 😊</strong></blockquote><blockquote>Join Coinmonks<a href="https://t.me/coincodecap"> Telegram Channel</a> and<a href="https://www.youtube.com/c/coinmonks/videos"> Youtube Channel</a> learn about crypto trading and investing</blockquote><h3>Also, Read</h3><ul><li><a href="https://coincodecap.com/best-crypto-to-invest-in-india-in-2021">Best Crypto to Invest in India</a> |<a href="https://coincodecap.com/wazirx-p2p"> WazirX P2P</a></li><li><a href="https://coincodecap.com/zero-fee-crypto-exchanges">7 Best Zero Fee Crypto Exchange Platforms</a></li><li><a href="https://coincodecap.com/best-online-casinos">Best Online Casinos</a> |<a href="https://medium.com/coinmonks/futures-trading-bots-5a282ccee3f5"> Futures Trading Bots</a></li><li><a href="https://coincodecap.com/what-are-decentralized-exchanges">Decentralized Exchanges</a> |<a href="https://coincodecap.com/bitbns-fip"> Bitbns FIP</a> |<a href="https://coincodecap.com/bingbon-review"> Bingbon Review</a></li><li><a href="https://coincodecap.com/buy-crypto-with-credit-card">10 Best Places to Buy Crypto with Credit Card</a></li><li><a href="https://coincodecap.com/5-best-crypto-trading-bots-in-canada">Best Crypto Trading bots in Canada</a> |<a href="https://coincodecap.com/bybit-binance-moonxbt"> Bybit vs Binance</a></li><li><a href="https://coincodecap.com/best-crypto-exchanges-in-uae">5 Best Crypto Exchanges in UAE</a> |<a href="https://coincodecap.com/simpleswap-review"> SimpleSwap Review</a></li><li><a href="https://coincodecap.com/ways-to-buy-dogecoin">7 Best Ways to buy Dogecoin</a> |<a href="https://coincodecap.com/zebpay-review"> ZebPay Review</a></li><li><a href="https://coincodecap.com/futures-trading-signals">Best Futures Trading Signals</a> |<a href="https://coincodecap.com/liquid-exchange-review"> Liquid Exchange Review</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c958467805a3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/coinmonks/crypto-and-our-society-c958467805a3">Crypto and our society</a> was originally published in <a href="https://medium.com/coinmonks">Coinmonks</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>