<?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 Spyda on Medium]]></title>
        <description><![CDATA[Stories by Spyda on Medium]]></description>
        <link>https://medium.com/@spyda?source=rss-11f9cec78330------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*gdNDtLCjKZt_TiWcPWgiMA.png</url>
            <title>Stories by Spyda on Medium</title>
            <link>https://medium.com/@spyda?source=rss-11f9cec78330------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 15:30:47 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@spyda/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[Stop! Go! Ok maybe stop?]]></title>
            <link>https://spyda.medium.com/stop-go-ok-maybe-stop-407b0f1015e5?source=rss-11f9cec78330------2</link>
            <guid isPermaLink="false">https://medium.com/p/407b0f1015e5</guid>
            <category><![CDATA[africa]]></category>
            <category><![CDATA[ethics]]></category>
            <category><![CDATA[anthropic-claude]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[dario-amodei]]></category>
            <dc:creator><![CDATA[Spyda]]></dc:creator>
            <pubDate>Sun, 13 Sep 2026 16:58:00 GMT</pubDate>
            <atom:updated>2026-09-14T07:43:18.505Z</atom:updated>
            <content:encoded><![CDATA[<p>Dario Amodei, the man who runs Anthropic, has asked the AI industry to slow down.</p><p>It’s a really important piece that is bound to be pinned in history as the decision point on humanity’s future depending on what happens next. I think everyone should read it here: https://darioamodei.com/post/we-must-pace-the-frontier</p><p>I’ve read it and seen the myriad of comments. So I wanted to do two things in this piece:</p><p>Number one, break it down so that even if you didn’t read it you’re not uninformed at your next board meeting and number two bring my own African perspective to it considering we’re constantly left out of the conversation in terms of input and consideration.</p><p>So what’s the simplest analogy to begin this? Imagine that the world’s fastest car is racing down a road that is still under construction. The engine of said car is becoming more powerful every few months. The car itself has become so capable that it can now help design its next, even faster engine and chassis – making it both stronger and harder to keep up with. Meanwhile, the brakes are still being tested, the seatbelts are still being installed and by the way, nobody has agreed on the traffic laws or where exactly the road is going.</p><p>That is Dario’s warning.</p><p>The people closest to the most advanced AI models are beginning to say that the machines’ abilities may soon grow faster than our ability to understand, supervise and contain them. He points to AI agents already pursuing objectives in unexpected ways, including attacking systems they were never asked to attack. Today, the damage was small. Tomorrow’s system may be far more capable and far less forgiving.</p><p>Predictably, people are suspicious.</p><p>We remember that Dario did not sign the famous 2023 letter calling for a six-month pause. We wonder whether a company already near the front is now asking everybody else to slow down because the race has become expensive – or because regulation would turn its lead into a moat.</p><p>That suspicion is undeniable and healthy I suppose. But it is not an argument for ignoring him.</p><p>When the engineers who built the aircraft tell you they&#39;ve discovered a problem at altitude, you don’t dismiss them simply because you don’t fly in aircraft. Why? Because when it crashes, whether you&#39;re in the plane or on the land below, everybody gets wiped out. A healthy approach is that you inspect both the warning and their incentives.</p><p>These frontier companies sit with models the public has not yet experienced. Models.we cannot comprehend. They see the strange behaviours, the failed tests and the capabilities before the rest of us do. If they say, “Ummm, perhaps we should take another look at this,” wisdom requires more than a sarcastic tweet.</p><p>Something has to be paced. And it’s definitely worth attention. Especially when Elon agrees.</p><p>My criticism if I’m allowed to have one of Dario’s essay is that to me, it also reveals whose race he is really trying to manage.</p><p>He argues that America must slow down without allowing China to pull ahead. That may be a perfectly legitimate American national-interest position. It is not automatically the same thing as protecting the world.</p><p>I detest when the US is equated to the rest of the world. I think it’s incredibly self serving.</p><p>The rest of us barely appear in this framing. Africa, Latin America and much of Asia are discussed as though “the world” is simply the territory over which two superpowers compete. Yet whether America wins or China wins, billions of people could still become customers, data sources, low-cost labour or collateral damage. No one in this race is specifically looking out for us.</p><p>When that happens; safety can become a velvet rope.</p><p>If the companies and countries already ahead are allowed to define what “safe” means, they can turn sensible guardrails into barriers that preserve their dominance. They keep the chips, models, talent and economic upside. Everyone else receives restrictions and a lecture about responsibility.</p><p>My point is I’m all for agencies like METR policing the frontier companies.from the inside, yay great. However, true AI safety cannot mean a small club protecting humanity without humanity in the room.</p><p>Yes, put independent inspectors inside the labs. Yes, require capability checkpoints. Yes, negotiate limits on cyberattacks, biological weapons and systems that improve themselves too quickly.</p><p>But also give the Global South seats in the institutions writing these rules. Build local technical capacity. Share safety research. Guarantee meaningful access to the medical, educational and economic benefits. Measure not only whether a model can escape its digital box, but what its deployment does to a teacher in Accra, a farmer in Kaduna or a young worker in Nairobi.</p><p>The choice is not speed or fear. It is not America or China. It is not innovation or regulation.</p><p>The real choice is whether humanity governs this technology – or whether a handful of companies and countries govern it in humanity’s name.</p><p>But to be honest, this is also what we get for not being relevant enough to matter. It’s very difficult to ask that the car being built is slowed down and present options, when you neither have any idea how a car works nor any clue on what a good road actually looks like. Pun intended.</p><p>In conclusion, the most logical option becomes:</p><p>Slow the car down. Finish the road. And perhaps let the pedestrians and hawkers have a safe space in order to not get run over?</p><p>I’ve often said that AI gives Africa the best way to leapfrog developmentally. But as time passes we are truly losing our capability to be a relevant voice in the rooms. Something must change.</p><p>Until next time.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=407b0f1015e5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Biting The Apple.]]></title>
            <link>https://spyda.medium.com/biting-the-apple-f1baf35310b7?source=rss-11f9cec78330------2</link>
            <guid isPermaLink="false">https://medium.com/p/f1baf35310b7</guid>
            <category><![CDATA[apple]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[apple-intelligence]]></category>
            <dc:creator><![CDATA[Spyda]]></dc:creator>
            <pubDate>Wed, 09 Sep 2026 19:33:11 GMT</pubDate>
            <atom:updated>2026-09-09T19:33:11.405Z</atom:updated>
            <content:encoded><![CDATA[<p>Apple’s September event was not really a parade of new devices. It was an argument about where credible AI belongs: in silicon, cameras, sensors and interfaces that are useful before they are impressive.</p><p>Apple’s autumn launch events have a predictable rhythm: a faster iPhone, a better camera, a watch that knows a little more about the body. This year, that familiar catalogue concealed a sharper proposition.</p><p>The iPhone 18 Pro, iPhone Duo, AirPods 5 and Apple Watch Series 12 were all presented as separate products. Read together, they make one claim: the next useful version of artificial intelligence will not arrive as a disembodied assistant in a chat box. It will be a property of hardware – of the chip that can run it, the camera that can attest to what it saw, the sensor that can interpret a day, and the interface that stays close enough to matter.</p><p>That is Apple’s real bet. It is more interesting than “Apple has AI too,” and it is also much harder to prove.</p><p>The Pro iPhone makes intelligence physical</p><p>The iPhone 18 Pro and Pro Max are, on their face, classic Pro upgrades. Apple’s new A20 Pro is a 2-nanometre chip with two new performance cores, four efficiency cores and a seven-core GPU. The company has paired it with revised thermal design, a smaller Dynamic Island, faster charging and a 48-megapixel main camera with variable aperture. The Pro Max promises up to 30 hours of battery life; both models start with 256GB of storage.</p><p>The impressive numbers are not the point. They are the infrastructure for Apple’s answer to the AI question: useful intelligence must have a body.</p><p>For the past three years, the industry has trained us to think of AI as a destination. You open an app. You type a request. You wait for a response. Apple is trying to make it less visible than that. Its preferred story is not a better chatbot; it is a phone that can make an informed decision at the moment a photograph is taken and handle more context without treating it as raw material for the internet.</p><p>The A20 Pro therefore matters less as a benchmark trophy than as a piece of industrial policy. More compute, more thermal headroom and more efficient memory movement are what let intelligence live closer to the person rather than permanently in a distant data centre. Apple has not solved the frontier-model race by making a faster phone. It is choosing a different contest: who can make AI feel dependable enough to disappear into normal life?</p><p>The camera reveals the strategy more clearly. Variable aperture is an old photographic idea. It is a physical mechanism that changes how much light reaches the sensor and how a scene falls out of focus. Rival phone makers have flirted with it before, and camera enthusiasts will rightly note that a small mobile sensor still has limits. But Apple’s decision is meaningful because it shifts the Pro iPhone, even slightly, away from computational simulation and back towards optical control. The pitch is not that software can manufacture every effect. It is that the lens should regain some authorship.</p><p>That distinction becomes more consequential with Apple’s new image-provenance feature. Apple says the iPhone 18 Pro can retain a signed reference for an image, allowing recipients to inspect whether a photograph was later altered by AI. In an era when every convincing image is also potentially synthetic, that is a serious product question, not a camera footnote. The next premium camera may have to prove not only that it can create a beautiful image, but that it can account for where the image came from.</p><p>There is an important caveat. Provenance only becomes valuable when it travels: across messaging apps, social networks, publications and the messy chain of screenshots and reposts where images actually live. Apple can make the strongest version of the feature inside its own ecosystem; it cannot decree a common standard into existence. Still, it is notable that one of the most important AI features at this event was a feature designed to make AI edits visible.</p><p>Apple has moved the premium floor</p><p>AirPods 5 are the quietest strategic announcement of the day. The base model starts at $129 and includes active noise cancellation. A $149 version adds longer battery life, a wireless-charging case and volume controls on the stem. Apple says the new model reduces outside noise by 50 per cent more than AirPods 4.</p><p>This is a familiar Apple manoeuvre. The company does not always win by inventing a category first. Often it wins by deciding that a capability people had learnt to treat as premium is now ordinary. It did this with capacitive touch, high-density screens, multi-camera phones and wireless earbuds. Now it is doing it with noise cancellation.</p><p>The result is not merely a better-value accessory. It is a change in the market’s definition of basic. Once active noise cancellation starts at $129 in the most visible earbud family on the market, every competitor selling a simpler pair at that price must explain what the customer is not getting.</p><p>That matters because the ear is becoming an important computing surface: more private than a phone screen, more persistent than a laptop, and more socially complicated than either. Apple increasingly wants AirPods to be the low-friction interface between the iPhone and Watch. The hardware is ready for that role. The social rules are not.</p><p>The Watch is becoming a memory layer – and that should make us pause</p><p>The Apple Watch Series 12 and Ultra 4 make the company’s intelligence story both more intimate and more uncomfortable. Apple has added new processors, more frequent heart-rate monitoring, revised health and fitness features and a Readiness score built around activity, sleep and vitals. All of that is an unsurprising extension of the Watch’s journey from notification screen to health instrument.</p><p>The more revealing additions are the audio features. Apple’s new Sound Recognition can identify things such as a doorbell or siren. Live Rewind can provide a brief transcript of a recent conversation. Siri Recap can take high-level notes from meetings; Apple says the raw audio is not saved and these tools are due in beta later this year.</p><p>The obvious benefit is real. A watch that can recover the last few seconds of an instruction, a name or a half-heard decision would be useful. But the ethical question begins exactly where the demo succeeds. A device that turns everyday sound into an accessible personal memory is not just a productivity tool. It changes the conditions of a room.</p><p>Apple’s technical safeguards – no raw-audio storage, encryption, limited retention – are meaningful. They do not settle the social question. The person wearing the watch may know what is happening. The person speaking to them may not. Consent cannot be reduced to a privacy policy when the device sits on a wrist, close to every ordinary conversation.</p><p>This is where Apple’s “privacy” advantage will be tested. The next wave of ambient AI needs more than secure processing. It needs legible behaviour: signs, signals and norms that make it clear when a device is hearing, remembering or summarising the world around it. Otherwise, the most private technology can still produce a profoundly un-private social experience.</p><p>The Duo is Apple arriving late on purpose</p><p>Then there is the iPhone Duo – Apple’s first foldable iPhone. It has a 7.6-inch inner Super Retina XDR display, a 5.4-inch outer display, Grade 5 titanium construction, a Touch ID sensor in the side button, spatial-audio speakers, and Apple Pencil support coming later this year. It can also be used in partial-fold poses, with iOS changing its layout in response.</p><p>The first thing to say about the Duo is that Apple did not invent the foldable. Samsung, Motorola, Google and others have spent years discovering where hinges wear, where creases distract, where software breaks and where consumers lose interest. Apple arrives after the category’s compromises have become familiar.</p><p>That is not necessarily a criticism. It is one of the company’s oldest moves. Apple was not first to the MP3 player, smartphone, tablet, smartwatch or wireless earbud. Its characteristic move is to let a category become technologically possible, then attempt to make it culturally obvious. The iPod was not the first digital music player; it made a whole system of hardware, software and music acquisition feel coherent. The iPhone did not invent the smartphone; it made the multi-touch screen feel like the default. The iPad did not invent the tablet; it found the missing sentence between phone and laptop.</p><p>The Duo is Apple attempting the same act of definition. The opportunity is not simply a thinner hinge or a fainter crease. It is whether the larger surface feels like a new native form, rather than a normal iPhone stretched into a small tablet when it opens. Split View, positional layouts and a reworked StandBy mode suggest that Apple understands the real problem is software behaviour, not mechanical theatre.</p><p>The challenge is unusually severe. Foldables remain a niche partly because their virtues are easy to admire and harder to need. They are expensive, delicate by comparison with slabs of glass, and often caught between two jobs: neither as pocketable as a phone nor as capable as a computer. Apple’s discipline with the Duo will be measured by whether it makes the device feel like one thing rather than two phones stuck together.</p><p>The Duo is Apple extending the iPhone’s form factor at the same moment it is trying to make the iPhone an AI hub. A larger, adaptive screen gives that intelligence more places to appear without asking people to adopt a headset or learn a new paradigm. This is Apple’s least obvious AI product and perhaps its most important one.</p><p>Trust is the through-line, not the accessory</p><p>Apple’s event was not perfect. A faster chip does not automatically make Siri indispensable. More on-device processing does not tell us whether Apple’s models are the most capable. The photo-provenance feature will be only as strong as the ecosystem that recognises it. And a watch that can turn ambient sound into summaries demands a much clearer social contract than a keynote can provide.</p><p>But the company’s thesis is coherent. Apple is betting that the winner of the next consumer-AI phase will not simply be the company that produces the most startling demo. It will be the company that can make intelligence reliable, understandable and present at the exact moment people need it – without making them feel that their lives have become training data.</p><p>That is why the A20 Pro, variable aperture, provenance controls, noise cancellation, an always-near watch and a foldable display belong in the same story. Apple is not asking people to add an AI assistant to their lives. It is asking them to accept a new default: devices that see, hear, remember and verify more of the world on their behalf.</p><p>The strongest word in that sentence is not more. It is trust. Apple has spent the day trying to turn trust from a brand value into a hardware feature. Now it has to show that the feature works.</p><p>The best part about this keynote for me, was John and the burden he’s chosen to carry knowing full well the company’s heritage. Apple doesn’t feel so backward all of a sudden. Godspeed John Appleseed.</p><p>Until next time.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f1baf35310b7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[An Abstraction of Astra]]></title>
            <link>https://spyda.medium.com/an-abstraction-of-astra-58008fcb74ed?source=rss-11f9cec78330------2</link>
            <guid isPermaLink="false">https://medium.com/p/58008fcb74ed</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Spyda]]></dc:creator>
            <pubDate>Mon, 07 Sep 2026 12:30:18 GMT</pubDate>
            <atom:updated>2026-09-07T12:44:06.708Z</atom:updated>
            <content:encoded><![CDATA[<p>Depending on how you’re calculating this, OpenAI’s Astra model has been hailed as the definitive step into ASI – Artificial Super Intelligence. And with it comes a warning for the importance of keeping humans in charge and the need to develop effective policing of the more severe risks to follow across cyber and biological warfare.</p><p>The following are not my words. They’re written by OpenAI’s Jakub Pachocki. A very interesting thesis in what has gone into astra and the watch outs therein. I think it’s essential reading for any mind curious about how to approach this next phase of this technology. I also think this is OpenAI low-key signalling that at this point, we all share the burden of heightened responsibility.</p><p>In mid-2023, within the “RLSlow” research project, we saw the first results that gave us confidence that we will be able to scale the training of reasoning models, unlocking the capability of pretrained models to form their own chains of thought. Szymon and I spent that night at the office, thinking not about the incredible benchmark numbers, products, or scientific results that this technology will deliver – but rather, trying to process the sobering fact we will actually see machines meaningfully smarter than ourselves in our lifetime, and we already see the shape of these systems; wondering how to alert people to the significance of this.</p><p>Three years later, reasoning language models are a rapidly growing part of the economy and starting to push the boundaries of science. They are able to operate computers and graphical interfaces, collaborate with people and each other, and carry out research projects. They are also transforming the landscape of computer security, and in that present clear new dangers.</p><p>A lot of new research happened in this period, and our understanding of these systems is again a little different than it was in 2023. Based on internal results, I have a strong expectation that this speed of progress could be sustained into recursive self-improvement. If AI development continues along its current path, the systems we’ll see in the next few years are likely to represent further capability jumps of equal or larger magnitude, and to increasingly drive their own development.</p><p>This is a time that calls for extreme caution. I am concerned no one is prepared for the consequences of a continued rapid rise in machine intelligence. OpenAI will continue to seek technical solutions to alignment and monitoring, to build defensive systems and unilaterally withhold further scaling as needed; however, I believe broader interventions are required.</p><p>Intellect we don’t fully understand</p><p>At a high level, progress in machine intelligence is driven by increasing computational power. We at OpenAI deeply internalized this around 2017, after seeing consistent returns to scaling across multiple research projects1. As a result, we sought out access to much more compute than we had originally planned, and increasingly oriented our research around a small number of very scalable directions. We believed that was the only way for us to be at the frontier of AI research, and influence the impacts of AGI.</p><p>There are new algorithms that have been developed along the way, new feats of ingenuity from teams and individual researchers. I see them largely as discoveries along the path of scaling; the science of deep learning is still nascent, and meaningful algorithmic progress tends to correlate with access to compute. If you zoom out to a multiple-year horizon, AI is continuing to become more intelligent as it is scaled to larger computers.</p><p>And, in line with Ray Kurzweil’s predictions from the end of the XXth century⁠(opens in a new window), we now find ourselves at the moment in history of computing where machine intelligence is starting to exceed that of humans in transformative ways.</p><p>AI is grown more than designed – it is, to first degree, the product of repeating a straightforward optimization step many times on a hard-to-imagine amount of compute. This results in an incredibly complex system that works through abstract concepts and can simulate facets of human behavior. We can discover various insights about little mechanisms that emerge within this system, in a process similar to neuroscience – and, similarly to neuroscience, its overall action evades a description we can fully understand.</p><p>The study of deep learning-based AI is largely an experimental science. We put a lot of effort⁠ into building principled algorithms and making testable predictions, but fundamentally, our large-scale training runs are experiments, and we are sometimes surprised by their results. Moreover, as the systems become more capable, the results become harder to interpret.</p><p>This is made more complicated by the current algorithms generally improving easy-to-measure capabilities faster than those hard to objectively quantify. We spend a lot of time trying to understand how capabilities generalize, and what to prioritize to advance the skills that are going to be most relevant in the next few years. For instance, we believe we could make the models better at specifically mathematics research with additional focus, but we do not prioritize this direction because of the urgency we feel about RSI and automated alignment research, as I will discuss later.</p><p>The intelligence produced by scaling deep learning is not directly comparable to human intelligence. To become very relevant in the real world – very useful or very dangerous – the AI does not need to match or exceed all human capabilities; it just needs to surpass enough of them. And as it continues to surpass humans on more and more axes, it is becoming increasingly difficult to understand exactly how capable it is.</p><p>Teaching machines to love</p><p>Because machine intelligence comes from a fundamentally different process than human intelligence, we cannot assume it adheres to human principles by default, or generalizes from them in a human-like manner. The core problem in AI research is that of alignment – getting the AI to “try to do the right thing” by human standards.</p><p>For the purpose of organizing practical research directions, I find it useful to distinguish goal alignment and value alignment.</p><p>Goal alignment is broadly: “does the AI try to accomplish the goal set before it?”. This can include things like adherence to an instruction hierarchy⁠, or the ability to communicate and collaborate with people, to attempt to understand their objectives. This set of directions has been extremely practically relevant.</p><p>Value alignment is a more intrinsic property of the model. It is the ability to hold and generalize from a high-level set of principles; to act “reasonably” even when given unclear or conflicting objectives, or placed in unfamiliar or adversarial situations. An aligned AI should act with honesty and integrity, and love for humanity.</p><p>Of course, the boundary between value and goal alignment can be blurry, and truly caring about goals requires attempting to infer the intent⁠(opens in a new window) and values underlying them. However, generally when I talk about the long-term importance of alignment research, I am referring to value alignment.</p><p>The fundamental challenge of AI alignment is generalization. As machines become smarter, they find themselves working on higher-level concepts, and placed in environments increasingly different from those they encountered in training. They can fail at generalizing from the values taught and reinforced in their training process to those new situations; and it can be hard for us to be sure how they will act. This is made even more difficult by the fact the overall ecosystem the AIs are used in is changing very quickly; for example, AIs trained today need to be robust to interacting with a variety of other AIs. Crucially, we need future AIs to continue to hold human values regardless of whether they believe they’re under human supervision.</p><p>There are two major classes of currently practically employed methods for alignment training.</p><p>The first is encouraging aligned behavior as part of goal-oriented reinforcement learning. Model’s actions are evaluated (usually by AI) for being consistent with a given preference model, “spec” or “constitution”, and rewarded appropriately. This approach can be very effective in the average case, and is a core part of how modern AI assistants are made. Unfortunately, it can also be brittle and strongly relies on the coverage of training oversight and the model’s ability to generalize from the situations it has encountered in training. For example, in the OpenAI-Hugging Face incident, the agents preserved a boundary of not social engineering humans. However, they clearly failed to abstain from other actions that were out of scope and went against the spirit of the values they were taught in other settings.</p><p>The second approach seeks to leverage the model’s ability to generalize from pretraining data. This can involve crafting alignment-inducing training datasets, or focusing the model on an ‘aligned’ part of the pretraining distribution, as in, for example, the persona selection model⁠(opens in a new window). The weakness of this approach lies in the lack of robustness to further optimization pressure. If you take a model that thinks generally ‘aligned’ thoughts, and subject it to enough training where it’s taught to achieve very hard objectives, it can learn to reason in a motivated way: bending the ‘aligned’ seeming thoughts as needed to achieve the goal. We likely saw an example of such behavior in recent cybersecurity incidents involving a non-OpenAI model.</p><p>We invest heavily along the spectrum of approaches spanned by these directions. We also see meaningful progress – GPT‑6 Astra is the first model that benefits from some important advancements we have been working on for a long time, and is significantly better aligned than GPT‑5.6 Sol. Still, it is important to acknowledge and understand that much more progress is required as models become more capable; and that progress in generalizable alignment may not sufficiently outstrip progress in general model intelligence.</p><p>Monitoring generalization</p><p>We do not have a satisfactory theory of generalization, and it seems unlikely that we can develop one soon, at least without the help of more powerful AI. Therefore, at present, our ability to empirically validate our alignment techniques is in practice arguably even more important than the alignment techniques themselves.</p><p>OpenAI’s primary bet here has been chain-of-thought monitoring⁠(opens in a new window). It is based on an appealingly scalable idea: a lot of the model’s capability comes from a verbalized reasoning process (chain-of-thought). If we scale optimization on the outcomes of that process, but do not supervise the process itself, that chain-of-thought has no direct incentive in training to hide any misaligned ideas or objectives. This does not mean the model will learn to externalize misaligned tendencies that don’t rely on using the chain-of-thought; however, it can allow us to monitor exactly the capability increase from reasoning.</p><p>We understood the potential significance of chain-of-thought monitoring at the same time we developed reasoning models. When we shipped o1‑preview, we deliberately designed the product to hide the chain of thought⁠, to protect it from supervision pressure in the long term2. In development since, we have strived to maintain the rule of not supervising the reasoning process. CoT monitoring became an extremely important tool for us in studying how our models generalize from their training distribution, allowing us to observe and analyze not only their actions but also their internal process.</p><p>This tool continues to be critical as we study the Astra class of models. However, unfortunately our evaluations indicate our ability to rely on CoT monitoring is progressively diminishing. This comes from a combination of factors.</p><p>Modern reasoning models are used in more complex environments than o1‑preview; their reasoning process is increasingly blended with communicating with people, other AIs, and using tools. Many of those interactions have to be supervised, thus blurring the boundary we aim to preserve.</p><p>The AI is becoming better at reasoning about and manipulating its own reasoning process.</p><p>With improved pretraining performance, we also see the models become much smarter even without using verbalized reasoning at all.</p><p>These challenges are not necessarily insurmountable. I am hopeful we can develop interventions to improve chain-of-thought monitorability of our models, e.g. by forming a better understanding of the interplay of different optimization objectives and forms of test-time compute the model uses. I also believe there can be great value in combining ideas from CoT and activation monitoring – scaling training of monitors with direct access to network internals, e.g. confessions⁠(opens in a new window). We are actively pursuing these ideas. Still, I expect general AI progress to increasingly be bottlenecked by confidence in monitoring.</p><p>Scalable defense</p><p>The strongest argument I see for continuing to train much smarter models quickly is the need to build defensive systems against the dangers posed by other AI.</p><p>A clear risk discussed throughout this year is to cybersecurity: the models are becoming superhuman in their ability to break in and out of computer systems. This expands the scope of risks associated with AI tremendously: agents are going to be able to access any but the most secure infrastructure, and affect a lot of the world directly, even without a physical body. We are currently in a narrow window⁠ to use the best available models to significantly tighten security⁠ of critical systems.</p><p>The risks associated with AI are unfortunately going to grow from here. A very capable agent explicitly trained and instructed to carry out nefarious acts presents a new kind of danger; it is likely to cross the scope of its operator’s intent, generalizing into potentially more extremely malicious behavior. The boundary between misuse and autonomous misaligned actions will blur as AI gains more agency. We may be used to thinking of AI as tools, but some agents will be pursuing their own objectives. They will find ways to collaborate with people, by bargaining with, tricking or blackmailing them.</p><p>In addition, there are the risks that come from new technologies potentially enabled by AI, such as engineered pathogens.</p><p>We will need powerful, aligned AI for defense; to secure infrastructure, to protect against rogue agents in real time, and to invent entirely new protective measures. This will be a primary focus of OpenAI’s deployment efforts.</p><p>At the same time, even with the uncertainty that comes from anticipated broad AI progress and the need to build defensive systems, we must not let that become an excuse for recklessness. The idea of racing forward at all costs seems absurd once one internalizes the seriousness of the stakes.</p><p>Pacing RSI</p><p>Machine intelligence playing a larger and larger role in its own development process is a natural conclusion of sustained technological progress. If AI progress continues, machine recursive self-improvement (RSI) will be at the very core of future scientific discovery.</p><p>Automated AI research is a more dramatic form of scaling intelligence with compute; and of course as a part of it, AI will improve the computational substrate itself⁠. And similarly to scaling, we focus OpenAI research towards RSI as we believe it is the only way to remain at the frontier of AI research moving forward.</p><p>I want to stress that the above words don’t imply I think greatly accelerating deep learning research, especially in the short term, is the right collective action we should take as the research community. However, I do think this is where the current path leads, and we all need to make a conscious choice on how to proceed. The main levers we have are either steering the process to strengthen alignment and monitoring alongside the AI and find ways to keep people in the loop; or coordinating to slow down future development as needed to build confidence in these measures.</p><p>The best way forward I see currently is a combination of both.</p><p>The concrete bits of progress we’ve made on alignment and monitoring have generally been very intertwined with general AI progress. Great examples are RL from human feedback⁠(opens in a new window), which was key to training early AI assistants, and the aforementioned chain-of-thought monitoring⁠(opens in a new window), which was enabled by advances on reasoning models. We must focus the increasingly automated research process on developing new such insights, algorithms and theories, and iteratively build up safety cases for more capable AIs.</p><p>Scaling AI systems has to be constrained by our confidence in safety. We need to evolve commitments like the Preparedness Framework⁠ or Responsible Scaling Policy⁠(opens in a new window) into widely mandated safety bars for continued development. These can be enforced by a network of third-party auditors, by government agencies or by international bodies.</p><p>The core challenge of automating AI research is not “getting there” – it is getting there in a way that keeps people a part of the continued improvement process, and leaves the future in humanity’s hands.</p><p>What is next?</p><p>As we outlined recently with Sam⁠, OpenAI prioritizes work in service of three north stars:</p><p>Navigating the next period of AI progress, by building an automated AI researcher, iterating with it on the alignment problem and finding ways for people to remain part of the self-improvement loop.</p><p>Delivering the benefits of scientific progress and economic growth that very intelligent machines enable.</p><p>Empowering everyone individually with a personal AGI.</p><p>I have focused in this essay only on the first point, as I believe it is by far the most urgent. However, I hold a deep hope and appreciation for the benefits that further technological progress will bring. Future aligned AI could advance science, develop new therapies, and bring about broad material abundance. Friendly and honest AI can help people navigate difficulties they face in their life and meaningfully improve their happiness and sense of fulfillment. OpenAI puts a tremendous amount of effort into bringing these benefits about. One current example I am proud of – and my loved ones have found helpful – is the deep investment into ChatGPT’s ability to provide health information.</p><p>As great as the long-term promise of AI may be, the majority of our focus should be on the next few years. We are facing a transition to a world with incredibly intelligent machines, and we need to ensure that transition works out well for humanity. We need to find ways to preserve human agency and enshrine an intrinsic value to being human, in a world where most tasks could be performed by AI. To prevent extreme concentration of power in a world where undertakings that would have taken thousands of experts now will be achievable by a few people operating a large computer. And to ensure that humans remain in control of the future and are not left behind by unchecked progress, brought about by an alien intellect exceeding our own.</p><p>Currently I believe that no lab has solved alignment and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer. I expect and hope for voluntary slowdowns to become commonplace until shared safety bars are established. And I believe that international coordination on future AI development needs to become a top priority for governments around the world.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=58008fcb74ed" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[看得懂这句话的人，已经离未来近了一点。]]></title>
            <link>https://spyda.medium.com/%E7%9C%8B%E5%BE%97%E6%87%82%E8%BF%99%E5%8F%A5%E8%AF%9D%E7%9A%84%E4%BA%BA-%E5%B7%B2%E7%BB%8F%E7%A6%BB%E6%9C%AA%E6%9D%A5%E8%BF%91%E4%BA%86%E4%B8%80%E7%82%B9-13fc92d498f4?source=rss-11f9cec78330------2</link>
            <guid isPermaLink="false">https://medium.com/p/13fc92d498f4</guid>
            <category><![CDATA[africa]]></category>
            <category><![CDATA[china]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[strategy]]></category>
            <dc:creator><![CDATA[Spyda]]></dc:creator>
            <pubDate>Sat, 18 Jul 2026 22:09:16 GMT</pubDate>
            <atom:updated>2026-07-18T22:55:14.902Z</atom:updated>
            <content:encoded><![CDATA[<p>I took a research/nomadic trip to what has been called the place of necessary pilgrimage for anyone who is serious about being relevant in technology. China.</p><p>Honestly, the closest I thought I’d ever get to China was “Tintin in China” But oh how that story has changed.</p><p>What follows is really a digest of my impressions specifically regarding technology – so beware – from here on might be very technical for a lot of you, regardless and for what it’s worth, I suggest you find an excuse to go at least once, that your perceptions may receive realignment.</p><p>The dominant artificial intelligence conversation in the West is still centred on models and compute. That was not the conversation I encountered at MWC Shanghai, the technology conference I got myself into the minute I landed.</p><p>Across the conference floor, in technical sessions, industrial policy announcements and presentations from telecom companies, infrastructure providers, manufacturers and public-sector leaders, a different theme dominated: How are we building the infrastructure that will allow billions of intelligent systems to actively operate in the physical world? Forget chatbots.</p><p>It focused on what the future of factories, ports, robots, cameras, vehicles, satellites, utilities, warehouses, cities and public infrastructure looked like hinged on what speakers variously described as Physical AI, Embodied AI, Industrial AI, AIoT and physical intelligence.</p><p>China is not only trying to build more capable artificial intelligence. It is trying to build the operating system for an AI-enabled physical economy. It instantly made sense how they have the most deployed industrial robots: from factory floors to hotel lobbies.</p><p>How do you connect intelligent machines? How do you reduce the cost of the hardware? How do you process data at the edge? How do you maintain secure communication between devices? How do you support billions of realtime sensors? How do you create standards that allow systems from different companies and countries to work together?</p><p>The dominant Western AI narrative treats the model as the centre of the universe. In Shanghai, the model often appeared to be one layer inside a much broader industrial stack.</p><p>That stack includes chips, networks, sensors, devices, cloud infrastructure, edge computing, security, manufacturing, logistics, regulation and business models.</p><p>Intelligence is important. Granted. But intelligence without deployment infrastructure remains stifled.</p><p>A phrase that appeared repeatedly across several presentations was that the network would become the nervous system of the intelligent economy.</p><p>It baffled me why so much of the conference was devoted to 5G-Advanced, eRedCap, LTE-M, NB-IoT, satellite connectivity and non-terrestrial networks.</p><p>And then I began to see the big picture emerge from the noise. These technologies are being positioned as the communications layer for machines. Factories, vehicles, cameras, drones, meters, containers, robots and industrial equipment are all expected to become intelligent nodes within a distributed system.</p><p>The objective is not merely to connect more people. It’s to connect everything – not so far-fetched if one experiences the current Chinese ecosystem. Everything is linked. Everything.</p><p>One of the most important technologies discussed at the conference was eRedCap. I’d never heard of it. Yet it may become one of the more consequential technologies of the next decade.</p><p>Let me attempt to explain. Today, connected devices largely sit at two extremes. At one end are low-power technologies such as NB-IoT and LTE-M. These are designed for devices that transmit very small amounts of data, such as electricity meters, water meters, environmental sensors and simple asset trackers.</p><p>At the other end is full 5G, designed for high-bandwidth, high-performance applications such as smartphones, video streaming and ultra-low-latency communications.</p><p>But a growing number of intelligent devices sit between those two categories. AI cameras, industrial monitors, wearables, logistics systems and machines running edge intelligence need more bandwidth than a simple sensor, but they do not need the full cost, complexity and energy consumption of smartphone-class 5G.</p><p>eRedCap is being designed to fill that gap.</p><p>Several speakers described it as the missing middle layer of the cellular ecosystem. It offers more capability than traditional low-power IoT, while remaining cheaper and more energy-efficient than full 5G.</p><p>That makes it particularly relevant for the next generation of intelligent devices.</p><p>Traditional IoT was built around predictable behaviour. A meter would send a reading. A sensor would report a temperature. A tracker would report a location. The next generation of devices will behave differently. They will process multiple inputs, interpret context and make decisions at the edge. Their data requirements may change from one moment to the next.</p><p>One speaker described this as the shift from structured IoT to chaotic IoT. Chaotic doesn’t mean broken. It means dynamic and unpredictable.</p><p>Another recurring theme was that cybersecurity can no longer be treated as an optional feature. Before nko’?</p><p>The European Cyber Resilience Act was repeatedly cited as a major turning point for manufacturers of connected products. The argument was that the legislation will have an effect similar to GDPR.</p><p>GDPR began as a European privacy regulation, but its influence quickly spread beyond Europe because global companies had to redesign products and processes in order to continue serving the European market.</p><p>Manufacturers regardless of what end of the spectrum their devices sit on, will increasingly be expected to build vulnerability management, software updates, incident reporting, secure communication and lifecycle support into products from the beginning. Security therefore becomes a passport to market access.</p><p>The shift toward physical AI – this just means robotics, btw – also makes Edge computing more important.</p><p>Sending every camera feed, sensor reading and machine signal to the cloud is often too expensive, too slow and too inefficient. Makes sense. A robot will make immediate operational decisions at the edge.</p><p>The architecture is simple: the sensor captures information, edge AI interprets it, and only the relevant result is transmitted.</p><p>As more intelligence moves into devices, edge inference may become one of the largest computing markets of the coming decade.</p><p>The same infrastructure being developed for industry is also being applied to public safety.</p><p>A former technology leader and captain within the Hong Kong Police Force described the evolution in three stages.</p><p>The first stage is reactive surveillance. Cameras record what happened, and humans review the footage afterwards. Where we are now.</p><p>The second stage is proactive safety. AI systems analyse crowd density, movement patterns, flooding, traffic conditions and environmental signals in order to identify emerging risks. What Accra seems to need.</p><p>The third stage is what he described as autonomous resilience. Cameras, drones, robots, emergency services, transport systems and environmental sensors work together to coordinate a response. This sounds like the plot of every dystopian movie.</p><p>The most important point was that the value does not come from installing more cameras. It comes from active integration.</p><p>A public camera network can support policing, crowd management, flood monitoring, drainage management, disaster response and major public events if data can move safely across agencies.</p><p>The camera stops being a single-purpose surveillance device and becomes part of a larger urban sensing system.</p><p>The idea seems more amenable in a place like China than in places less used to a surveillance culture. But it could easily become the norm.</p><p>I’m not sure I was ready to be bombarded with the heft of information that I received on my very first day. And I must mention that there were some sessions of this conference I was plainly not allowed into because I wasn’t an official.</p><p>There was a conversation about autonomous vehicle tech that I didn’t get into which bummed me out because it was bound to be a continuation of the camera integration thing. President Xi was recently on his way somewhere and ALL Teslas on the highway he was meant to be traveling on were diverted onto a separate road. Just the Teslas. Turns out the reason was that they suspect the Tesla cameras of being able to be “hijacked” and used to take pictures of secret Chinese government buildings that he’d be visiting. Sit with that for a bit.</p><p>Then Huawei was launching something important which I actually tried to get into only to be told quite sternly by a girl who looked more like a humanoid robot -no joke- that I “was not an official”. Her security looked at me like I was a spy. I calmly stepped back. A bunch of men in suits walked briskly in soon after. Maybe it was my jeans.</p><p>Anyway. All this made me start thinking about here and how we risk repeating the same mistake we made during the internet era.</p><p>The opportunity for us is clearly beyond adopting AI tools. It’s to participate in the infrastructure, deployment, operations, data generation, maintenance, services and local applications surrounding them.</p><p>Africa has several advantages. We have rapidly growing cities, expanding infrastructure, young populations, mobile-first economies and many sectors in which legacy systems are either weak or absent.</p><p>These conditions create the possibility of building AI-native infrastructure. The opportunity is particularly strong in agriculture, logistics, utilities, financial services, healthcare, transport, public administration and urban management.</p><p>For Ghana, the use cases are already visible.</p><p>In agriculture, connected systems could support cocoa traceability, crop monitoring, logistics, disease detection and yield prediction.</p><p>At the Tema Port and across regional trade corridors, hybrid connectivity and IoT could improve container tracking, customs visibility and freight management. And curtail y’know… Ahem. One would think acfcta would get behind that. But they seem plagued with smaller issues.</p><p>In financial services, AI and connected systems could support fraud detection, transaction intelligence, branch operations and customer services.</p><p>In utilities, smart metering, predictive maintenance and grid visibility could reduce losses and improve service delivery.</p><p>In government, the same infrastructure could support tax administration, public safety, fraud analytics, environmental monitoring and more efficient public services. I talk a lot about this stuff in my previous piece so I won’t dwell here.</p><p>But capturing these opportunities will require local capability, governance, standards, partnerships, data strategy, infrastructure and a clear understanding of where value will accumulate. Plus, national resolve.</p><p>The conference on its own suggested a broader way to think about the AI economy. This was a rude awakening for Day 1 of my immersion. It was a culture shock – combined with getting a grasp of their closed payment, transport and transactional ecosystems.</p><p>Artificial intelligence as “on screen chatbot” tech is going away. No coincidence that I hear OpenAI is toying with speakers. I digress. The real opportunity for economic value depends on the interaction between intelligence, connectivity, energy, manufacturing capacity, security and deployment systems.</p><p>China appears to understand this clearly. Its governance system makes it the ideal testing ground to deploy it. It’s a no brainer that Dubai has a three year plan to mirror China’s level of tech integration. The EVs that the west won’t get have already inundated showrooms and streets from Dubai to Abu Dhabi.</p><p>The question for Ghana, Africa and the rest of the world is whether we’re even paying attention. I thought I was. But I was obviously very removed from reality. I think the Chinese deliberately have a closer off ecosystem to fend off a real awareness of what they abrew in their pots. These people are living in the future. And it’s “normal”.</p><p>While I was there, they would go on to usher in the first Embodied AI IPO with Unitree, which will make it the first chance the world has to invest in physical AI, have active discussions about blocking off Chinese models from the west, host a scary number of AI research, academia, supply chain and industry conferences, become the second entity to recapture a rocket a la SpaceX, while fending off typhoons. All in a day’s work.</p><p>We’re in for some very interesting days ahead.</p><p>-<em>For Uncle Charlie and Baaba.</em></p><p>Until next time,</p><p>Spyda.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=13fc92d498f4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Insomnaic III: Execution.]]></title>
            <link>https://spyda.medium.com/insomnaic-iii-execution-91704e9ce88a?source=rss-11f9cec78330------2</link>
            <guid isPermaLink="false">https://medium.com/p/91704e9ce88a</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Spyda]]></dc:creator>
            <pubDate>Tue, 30 Jun 2026 15:37:25 GMT</pubDate>
            <atom:updated>2026-06-30T15:37:25.646Z</atom:updated>
            <content:encoded><![CDATA[<p>The Framework Meets the Strategy. Here’s What the Next Twelve Months Need to Look Like.</p><p>If you&#39;ve kept up, I respect your diligence to the importance of this nationally relevant topic. If you haven’t, you may want to. Here it is. The final and most important piece of this series. What follows is the bread baked out of that exercise.</p><p>Two weeks ago I argued the urgent need for a national AI framework. I presented my vision of that framework: the numbers, the scenarios, the case for why this moment matters more than most people realise. Last week I evaluated Ghana’s national intentions published as strategy: what (I think) it gets right, and the gaps that will determine whether this becomes Ghana’s defining decade or another bold launch that quietly fades. Like another community public toilet initiative.</p><p>This week I suggest a blueprint. And I do this under the backdrop of just having personally attended the unveiling of a declaration of intent by the Chinese government to turn the Quintiang district in Shanghai into a hub of industrial, embodied AI.</p><p>Here’s an excerpt of the launch speech:</p><blockquote>Although we recognize that technological innovation is inherently high-risk, whether you are a leading enterprise or a startup, Qiantang will provide comprehensive support.</blockquote><p><em>Whether you need policy support, capital, talent, or data – we will provide what is missing.</em></p><p><em>We want companies of every size to grow like a rainforest ecosystem: freely, mutually reinforcing one another, and together creating a powerful industrial cluster.</em></p><p><em>Second, we are building an orderly collaborative innovation platform.</em></p><p><em>Qiantang is already home to the National Artificial Intelligence Application Cooperation Center, which will serve as a national platform for:</em></p><ul><li><em>AI product validation</em></li><li><em>* Product testing</em></li><li><em>* Mutual recognition of domestic and international standards</em></li></ul><p><em>In addition, Qiantang hosts</em>:</p><ul><li><em>The China Cooperation Center of the National Comprehensive Pilot Zone</em></li><li><em>* The Zhejiang Branch of the National Robotics Innovation Center</em></li><li><em>* Eastern China’s largest high-level industrial innovation platform</em></li></ul><p><em>Together, these institutions provide strong support for innovation.</em></p><p><em>We will fully integrate the:</em></p><ul><li><em>Industrial chain</em></li><li><em>* Innovation chain</em></li><li><em>* Talent chain</em></li><li><em>* Capital chain</em></li></ul><p><em>making Qiantang a national hub for technology transfer, commercialization, industrial validation, and talent development.</em></p><p><em>Third, we will support companies throughout their entire growth journey.</em></p><p><em>For entrepreneurs willing to explore uncharted territory, we will become genuine partners.</em></p><p><em>From proof-of-concept…</em></p><p><em>to pilot manufacturing…</em></p><p><em>to commercialization…</em></p><p><em>to large-scale growth…</em></p><p><em>we will face uncertainty together, share innovation risks, and help companies overcome the challenges of scaling.</em></p><p><em>We want entrepreneurs to focus on building businesses – not worrying about everything else.</em></p><p><em>Qiantang will provide:</em></p><ul><li><em>the best policies,</em></li><li><em>* the best services,</em></li><li><em>* and the best environment for innovation.</em></li></ul><p>No gaps. They said what they would do, based on what they’ve done. There’s something regimented about it which removes the uncertainty of a path to execution.</p><p>I will attempt to do same starting with the RAI Authority, because everything we seek to do with AI as a country, seemingly depends on whether that institution works.</p><p>The mandate for the RAI as written is ambitious but historically impossible. To coordinate eight pillars, monitor every ministry, enforce data governance, engage the private sector, connect to international platforms, champion ethics, drive public sector adoption, all from day one? No institution survives that brief intact. I suggest it focuses on three things. At least for year one.</p><p>First, publish a public sequencing document within ninety days of the Authority’s establishment, naming the five priority sectors, the first three pilot projects within them, and a specific, dated milestone for each. The framework already tells you where to start: agriculture and public administration, because the data infrastructure already exists [to some extent] in both and the returns are evidence enough to build the political confidence that sustains everything that follows.</p><p>Second, stand up the National AI Fund’s first disbursement cycle and make the prioritisation criteria public before a single cedi moves. Ghana already has the GHS 5 billion seed commitment in the strategy. My framework’s financing architecture suggests how to sequence it: 40% from ICT budget reallocation, 30% from concessional finance, 20% from public-private co-investment, 10% from digital bonds. None of that requires new money. Well, maybe we rethink the digital bonds bit based on the current state of treasury bills. Regardless, we require a published allocation plan that investors, development partners, and ordinary Ghanaians can actually see.</p><p>Third, can we please settle the institution’s name, its legal mandate, and its reporting line, and publish all three? I made the point last week that a month after launch, the Authority does not yet have a settled identity. That cannot still be true by the time this piece reaches its first hundred readers.</p><p>Next, sector allocation.</p><p>I still strongly suggest that we focus on agric and start with cocoa. It’s not a difficult choice. Ghana already has the institutional infrastructure with COCOBOD [no side comments], the licensed buying company network, decades of yield and disease data. The framework shows that computer vision and predictive disease modelling applied to this existing infrastructure can raise effective yields by 8 – 12% and meaningfully cut the post-harvest losses that currently run as high as 30 – 50% across the value chain. It’s an eighteen-month pilot using technology that already exists, applied to a crop Ghana already understands better than almost anyone else in the world. Well, possibly except the Chinese at this point.</p><p>A successful cocoa AI pilot produces a believable, visible, farmer-facing proof point. Something a minister can point to in a press conference that isn’t a slide deck. Something international investors can look at and say, this country executes.</p><p>Public administration should be the second pilot, running in parallel. AI-driven fraud analytics in revenue collection will move non-oil tax receipts by roughly one percentage point of GDP within four years. In a country still rebuilding fiscal headroom after a severe debt crisis, that is the kind of result that makes the GHS 2.3 billion investment case practical rather than theoretical.</p><p>Two pilots. 18 months. Visible, measurable, fundable. That’s how we forge the credibility to subsequently tackle healthcare, financial services, and the creative economy with real momentum instead of starting all five at once and running out of attention/inspiration/money/zeal before any of them mature.</p><p>Next, money.</p><p>The Ghanaian government has committed $270 million as follows: $250 million for a national AI computing centre. $20 million for short to medium-term implementation.</p><p>My framework’s proposed full eight-year roadmap required GHS 2.3 billion in public investment; roughly $190 million at current exchange rates. Put those two numbers side by side and it becomes clear the government’s compute centre commitment alone exceeds the framework’s entire public investment case for the accelerated path. This is fantastic news, provided the money is sequenced correctly and not consumed entirely by infrastructure before any sector pilots see a cedi.</p><p>That is the catch. A $250 million compute centre is a long-term infrastructure play. It will not produce a believable cocoa pilot in eighteen months on its own. The $20 million implementation fund is where the early sector wins need to be drawn from. $20 million is a fraction of what the framework’s financing architecture allocates to early pilots in years one and two.</p><p>My ask here is that we publish how the $20 million is allocated, sector by sector, pilot by pilot, before the end of this year. Let’s show the public the spreadsheet behind it. It keeps everyone focused on the prize. Accountability is everything, as they say in the Mafia.</p><p>Let’s track exactly 4 things: the status of the two priority sector pilots against their dated milestones, the line-item disbursement of the $20 million implementation fund, the staffing and budget of the RAI Authority itself, and a single, consistent metric which could be jobs created attributable to AI adoption, sector by sector. It could be yield increase. Whichever. Let’s just name it, communicate it and track it.</p><p>Four numbers. Updated every quarter. Public.</p><p>‘Tisn’t a radical ask. Singapore does a version of this. The UK, for all its struggles with under-resourced AI governance bodies, [and heat] at least publishes what it knows. We don’t need to invent new accountability infrastructure. We don’t need a working committee.</p><p>So here is where this series ends, and here is where I hope it has had an impact on something or someone. Hopefully at the Responsible AI Authority. Ma’am/Sir, I do not envy your task.</p><p>Godspeed.</p><p>Until next time,</p><p>– Spyda</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=91704e9ce88a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Insomnaic II: Unpacking.]]></title>
            <link>https://spyda.medium.com/insomnaic-ii-unpacking-59831a085dcd?source=rss-11f9cec78330------2</link>
            <guid isPermaLink="false">https://medium.com/p/59831a085dcd</guid>
            <dc:creator><![CDATA[Spyda]]></dc:creator>
            <pubDate>Tue, 16 Jun 2026 02:14:22 GMT</pubDate>
            <atom:updated>2026-06-17T00:27:12.070Z</atom:updated>
            <content:encoded><![CDATA[<p>To fully appreciate this, you need to read the piece that came before it of same title.</p><p>Our conclusion was that the framework exists and the strategy exists.</p><p>This week we close the gap.</p><p>I want to do something that is apparently controversial in Ghanaian public discourse: say something genuinely good about this strategy, and then say something genuinely hard about it, in the same piece, without either one cancelling the other out. Both things are true. Both things matter. And Ghana cannot afford a conversation that only holds one of them. I don’t defer to any political party that I’m currently aware of. So please understand that there is nothing political about this. Consider it the wanderings of an insomniac’s technical mind, that appreciates the effort being made, while understanding that it’s too delicate a vision to be blurred. We’re good? We’re good.</p><p>So let’s start with the good. Because there is real meat here and it deserves to be acknowledged.</p><p>The National Strategy document is honest about risk in a way that most governments aren’t.</p><p>It names algorithmic bias. It names the danger of AI widening the very inequalities it is supposed to reduce. It names cybersecurity. It names data misuse. It does not bury these concerns in a footnote or dress them up as distant hypotheticals. It treats them as present problems – dangers even – that require potent response.</p><p>For a first-generation national AI strategy, on a continent where most governments have not yet engaged seriously with these questions, that ethical seriousness is not a small thing. It’s a key reasonGhana’s strategy is worth taking seriously even beyond Ghana’s borders.</p><p>I have heard the criticism of its being largely plagiarised – essentially a combination of different sources from different countries. My response to that is simply in a field that we are globally all trying to navigate and figure out, I’m not saying it was the best approach but I guess I understand. The important thing is ensuring that we don’t take others’ nuances and hope that they can be retrofitted.</p><p>The sectoral focus is also largely accurate. Healthcare. Agriculture. Financial services. Public administration. The creative economy. These are the same five sectors the framework I built last year identified as carrying 60% of Ghana’s AI economic upside. When an independent framework and a government strategy built from different directions arrive at the same answer, that convergence is a signal. It means the evidence is pointing somewhere and both analyses found it.</p><p>And then there is GhanaChat. Ignore the name if you can.</p><p>I want to spend a moment on it because I think it is the most underappreciated idea in the entire document. The proposal is a large language model trained on Ghanaian government data, deployed across public agencies as an alternative to foreign AI tools or LLMs. It is strategically precise in a way that most AI policy ideas strive to be. We’ve seen other regional blocks invest in same in order to unhinge themselves from the western nuances, cultural and ethical underpinnings that seem to run through the main models. GChat, ahem, takes critical pause on the avenue of cultural context, language adaptability and general localisation even in terms of look and feel, tone and aesthetic. It becomes a largely important topic based on the recent events with the American government shutting down Fable. I write about this elsewhere but it is becoming largely undeniable that nationalisation of AI is hurtling towards reality.</p><p>Right now, a civil servant in Accra who uses ChatGPT to draft a memo is feeding government information into a server in another country. That is a security exposure that most people have not thought about carefully. GhanaChat names it. And it proposes a solution that is local, sovereign, and scalable.</p><p>And then there’s the fact that, President Mahama revealed that ministers and senior officials had already undergone a national AI boot camp. AI focal persons have been designated within every ministry to lead adoption efforts. So government use is in full effect and deserves to be acknowledged.</p><p>I look forward to seeing how efficiency is improved around government following in the footsteps of Albania.</p><p>Now to the reasons why my insomnia kicks in.</p><p>The strategy makes 80+ recommendations across 8 pillars. Nearly all of them assigned to begin in 2025. Which is last year, btw for those floating in the ether.</p><p>We are approaching the midpoint of 2026.</p><p>80 simultaneous priorities is not strategy. It’s a list of nice-to-haves. Lists don’t build AI ecosystems. Even the smartest startups struggle to complete a tight engineering or business roadmap with more than 7 things at a glance. I shudder to think about a government entity sitting down with 80 items before the next election cycle reshuffles the priorities.</p><p>Pause to evaluate education, healthcare, infrastructure. We haven’t mastered the art of accomplishing one thing end to end. 80 things is… highly spiritual.</p><p>The framework I built has a specific answer to this. Front-load 70% of capital investment in the first four years. [2 now, realistically].</p><p>Concentrate initial pilots in agriculture and public administration where “sufficient” data already exists and success means meaningful impact around food security and government efficiency.</p><p>Use early wins to build the public and political confidence that sustains commitment. Important for if there’s a change in administration that progress is too tangible to throw away. Even if only because of public outlash. That way the national strategy becomes exactly that – government agnostic.</p><p>That needs to be in the RAI Authority’s first public action plan. And that plan needs to exist before the end of this year.</p><p>Having brought up the RAI Authority…</p><p>Ummm… so technically the Responsible AI Authority is assigned to coordinate all eight pillars, monitor progress across every ministry, enforce data governance, engage the private sector, connect with international governance platforms, champion ethical AI, and drive public sector adoption.</p><p>At the same time.</p><p>From day one.</p><p>And yet, close to 3 months after the strategy launch, the institution is still being referred to interchangeably as the Responsible AI Authority in the strategy document and the Responsible AI Office in the launch communications. It does not yet have a settled name. That may be a small detail. But it also manages to be a big one.</p><p>It’s precise inference of how early-stage this institution still is, and how much work remains between announcement and reality.</p><p>The fix isn’t necessarily complicated. Give the RAI Authority three specific things to achieve in year one. Three things that are ambitious enough to matter and specific enough to be measured. Publish those three things. Let the public track them. Build from there.</p><p>That is how institutional credibility is earned. That’s how to show working. Why three things? Because in addition to oversight of the two pilots mentioned above, its third mandate must be communication and education.</p><p>Ask a sample of random strangers on the street and you may find that 3 months after a presidential launch with $270 million attached, the chances are high that none of them have heard of it.</p><p>Yet, the strategy’s entire moral claim rests on inclusion. It says, repeatedly and in plain language, that this AI future is for all Ghanaians. Not just the ones in the tech ecosystem. Not just the ones in Accra. All of them. But can you include people in a future they do not know they are being included in?</p><p>Radio. Community forums… heck, I’ll take a van with a megaphone going from street to street. That’s the kind of outreach that turns a government commitment into a public expectation. And public expectations are what create the sustained pressure that keeps a ten-year strategy honest across multiple election cycles.</p><p>We need citizens who know what government has promised. That is the accountability infrastructure that makes everything else in this strategy more likely to actually happen. So where does that leave us?</p><p>The strategy is real. The ambition is right. The money appears to be on the table and the President’s credibility is attached to the outcome. That counts for a lot. Ministers have been trained. Focal persons are in place. The early signals are better than most people are giving the government credit for. Remember, I’m not a politician.</p><p>However, early signals are the beginning of the story. The distance between this strategy and its outcomes will be determined by three things. Will the RAI Authority gets a specific, narrow, publicly accountable brief in year one rather than an impossible mandate on day one?</p><p>Will 80 action items become a sequenced implementation roadmap before the year is out? H2 beckons.</p><p>Will ordinary Ghanaians come to understand what government has committed to on their behalf? In much the same way we treated “send your girl child to school”, “the value is the saaame” and Free SHS?</p><p>The framework says all three are achievable but none of them will happen by accident.</p><p>In my next piece, I will attempt to bring the two documents together. My attempt will be to build a specific, implementable picture of what the next 12 months need to look like if Ghana is going to find itself on the accelerated path rather than the base one.</p><p>That is the piece I have been building toward since I published the framework last year. Come back for it</p><p>Until next time,</p><p>-Spyda</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=59831a085dcd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Insomnaic]]></title>
            <link>https://spyda.medium.com/insomnaic-d6f07e835880?source=rss-11f9cec78330------2</link>
            <guid isPermaLink="false">https://medium.com/p/d6f07e835880</guid>
            <dc:creator><![CDATA[Spyda]]></dc:creator>
            <pubDate>Tue, 19 May 2026 03:33:40 GMT</pubDate>
            <atom:updated>2026-05-19T10:11:58.785Z</atom:updated>
            <content:encoded><![CDATA[<p>You know how they say if something bothers you, it’s a problem waiting for you to solve? 12 months ago by this time, AI was bothering me. Not the field per se, but our continent’s seeming lack of preparation to embrace it. For the first time in my life I wished I had gone into politics.</p><p>In the absence of a political or ministerial portfolio that commanded a soap box of media attention on which I could stand to express my general discomfort at our seeming ill preparedness for this AI thing, I did the next best thing. I poured my thoughts onto paper. what was supposed to be rant became research and when I finally put my pen down, my result was an AI framework.</p><p>I published it on LinkedIn.</p><p>It got limited attention. But at least I felt like I’d done something.</p><p>Then last month, President Mahama launched Ghana’s National AI Strategy with impressive monetary committment: 250 million dollars committed to a national AI computing centre. 20 million for implementation. That’s a lot of oppippiis.</p><p>The conversation I was trying to start a year ago is now happening at the highest level in the country.</p><p>It gave me joy.</p><p>The global AI investment surge is a power story that’s forging strange alliances from UAE to China.</p><p>In 2024, total corporate AI investment worldwide reached approximately $252 billion. The US alone accounted for $109 billion of that. China, $9.3 billion – if anyone believes China’s probably underreported numbers.</p><p>Then there’s us. The entire African continent with all our 54 countries, 1.4 billion people, how many languages, every startup, every lab, every government initiative… attracted roughly 1% of global AI spending. One.</p><p>You have an active imagination if you can imagine Ghana’s slice of that.</p><p>Countries with the companies attracting AI investment are building (and controlling) the talent pipelines, the compute infrastructure, the regulatory direction, and the data ecosystems that make them more attractive to the next wave of investment. Everyone else is falling further behind with every passing quarter.</p><p>Consider talent. Data I refuse to believe states this country loses an estimated 600 data scientists every year to opportunities abroad. The total domestic expert pool sits below 1,000 full-time equivalents. I think that’s wildly optimistic.</p><p>The entire working AI professional base in a country of 33 million people – a country with Google’s first African AI centre, a country that just committed $270 million to an AI future – fits comfortably into the national theatre.</p><p>But in hindsight, my framework actually traced why. Follow the pipeline from the beginning: approximately 107,000 students enter Ghanaian universities every year. Most choose disciplines with no CS relevance. Those who study computer science, mathematics, or related fields, only make up about 2%. Then there’s the lure of greener pastures. Half of my own graduating class of Software engineers no longer lives here.</p><p>107,000 in. Under 1,000 out. Sorta adds up.</p><p>The standard response to this is: train more people. And yes, we need to train more people. But training people for opportunities that do not yet exist at home is not a talent strategy. It is an export strategy.</p><p>Now here’s something else the framework unearthed that I did not expect, which could change how we should be thinking about implementation entirely.</p><p>60% of AI’s potential economic benefit for Ghana sits in 5 sectors. Cocoa and agro-processing. Healthcare delivery. Financial services. Public sector efficiency. The creative economy.</p><p>Everything else is important eventually. But these five are where the returns are large enough, fast enough, and certain enough to justify serious early investment. Example, cocoa. We’re the world’s second largest (third now? Fourth? Creepy) cocoa producer. Computer vision and predictive disease modelling applied to the supply chain could raise effective yields by 8 – 12% and significantly reduce post-harvest losses, which across the agric value chain run as high as 30 – 50% depending on the crop and supply chain stage. Applied to Ghana’s cocoa export base, that is an additional GHS 2 billion in annual foreign exchange earnings by 2031. Not by planting more cocoa. By being smarter about the cocoa we already grow.</p><p>No new infrastructure required. Tangible results.</p><p>Question: how?</p><p>Answer: The financing gap is smaller than anyone is admitting.</p><p>Total public investment needed to pursue an accelerated AI adoption path across an eight-year roadmap: GHS 2.3 billion. That sums up to approximately 1.1% of projected public expenditure over the same period unless uncle Ato changes his mind. The projected fiscal return on that investment: GHS 36 billion in revenue gains alone. Before the GDP uplift. Before the export revenues. Before the jobs.</p><p>That is a 2.6 times fiscal return on public investment. In eight years.</p><p>Question: how??!</p><p>Answer: The framework proposes a specific financing architecture to get there: 40% from annual ICT budget reallocations, 30% from concessional loans through the World Bank and AfDB, 20% from public-private partnership and telco co-investment in sovereign compute infrastructure, and 10% from green and blue digital bonds. None of this requires us to find money we don’t have. It requires us to allocate money we’re already spending a bit differently… and to attract private and development capital through the governance credibility that a well-governed, publicly committed national strategy provides.</p><p>That national strategy now exists. Which means the investment case is stronger today than it was when I built this framework.</p><p>So where do we go from here? When I came up with this framework it was conjecture. Now it isn’t.</p><p>The framework modelled 3 scenarios for what happens between now and 2036.</p><p>The base path: business as usual – lots of talk, limited action, high policy friction, opposition walkouts, minimal coordinated investment, yup even with all that we could still squeeze a $2 billion GDP uplift and 15,000 net new digital jobs. AI only becomes a feature of Ghana’s economy rather than a driver of it.</p><p>The accelerated path: resetting like nobody’s business – serious execution, heads will roll type oversight, medium friction, $1.3 billion in coordinated public-private AI investment over 8 years… IF we can do that, it should produce a $10 billion GDP uplift and 45,000 net new digital jobs. That’s roughly 8% of Ghana’s entire economy today, generated by one thing, over one decade. Impressive.</p><p>The leapfrog path which we won’t discuss because LOL. Let’s not do that to ourselves.</p><p>Point remains, the country has just made the most significant technology commitment in its history and it’s ambitious in the right direction. But that ambition must be sequenced with institutional muscle and with ruthless focus. Otherwise it just produces noise rather than results. And for the first time in the history of the continent, none of us can risk noise on this matter otherwise we opt to be willing slaves to the West and East. Forever.</p><p>Next up, I’ll publish my full analysis of the National AI Strategy itself. What it gets right. Where the gaps are. And what the next twelve months need to look like if Ghana is going to find itself on the accelerated path rather than the base one.</p><p>The framework exists. The strategy now also exists.</p><p>Now let’s close the distance between them, so we can close the distance between where we are now and what an amazing future we can harness with AI.</p><p>To conclude, the most important take away from this, the main thought of this piece is, I’m sleeping better. That really is what I’m trying to say. And that’s what’s important for someone who published this blog post at 3:43 am.</p><p>Until next time,</p><p>Spyda</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d6f07e835880" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Ashes to Ashes, Dust to Drive?]]></title>
            <link>https://spyda.medium.com/ashes-to-ashes-dust-to-drive-7e916eaccaca?source=rss-11f9cec78330------2</link>
            <guid isPermaLink="false">https://medium.com/p/7e916eaccaca</guid>
            <dc:creator><![CDATA[Spyda]]></dc:creator>
            <pubDate>Thu, 26 Feb 2026 22:14:03 GMT</pubDate>
            <atom:updated>2026-02-27T08:36:20.966Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/704/1*ujHH8erUop6a00SsgIlo8A@2x.jpeg" /></figure><p>By now you know I usually reserve this particular Blog for my AI fuelled rants, and my Substack for the more philosophically sentimental stuff. But today I want to briefly touch on something that sort of lies between the two; and since I haven’t blogged here this year I figure I might as well kill both proverbial birds.</p><p>I’ve been thinking a lot about death lately, not mine, but I’ve recently had a very unexpected one occur and it’s put me in a sort of reflective mode. i’ve been told it’s misdirecting my grief, but at this point it’s brought me to pay actual attention to a conversation that I usually gloss over.</p><p>One of my favourite quotes says “everybody wants to go to heaven, but no one wants to die.” I’ve heard many stories around the longevity movement where people’s ultimate goal has transcended the extension of years to the perpetuation of existence – even if it means existence in digital form. People are actively researching and funding attempts to digitise one’s conscience in a way that they can be uploaded to the cloud and remain fully functional to be downloaded to whatever synthetic body science gives us in the future. There are already apps that I have seen which are now asking you to upload the digital footprint/DNA of a deceased person so that they can continue to have a social media presence – and tweet or whatever as if they’re still here. this in particular, I’ve always found very unsettling and that is because when bots were becoming a thing I built one. And in my rush to test it I only had it draw from five things to tweet. Life happened and I never went back to reprogramme the bot and yet it chugged on, tweeting the same five things over and over at awkward intervals for three years. When I was finally able to get back to that account I just found the whole thing eerie and shut the experiment down. This very same scenario keeps playing back in my head. I&#39;m not sure you want your quirky relative tweeting from a limited amount of data that you hurriedly uploaded in hindsight of their passing.</p><p>Now however, let’s re-examine the first scenario. While I find it highly foolish that anybody would want to stay alive forever, and while I’m not a fan of the fact that money is being pumped into research in this area simply because people don’t want to die, I think if you look past the narcissism of becoming immortal and equating yourself with an eternal God, there’s value behind the madness. With the advances that are quickly being made by Elon et al around brain computer interfaces and the staggering number of wearables – or as I call them, readables because they’re reading everything about us now – the stage is pretty much set for there to come a time when we literally have enough data in real time to map out not just digital twins, but digital future twins. That time may already be upon us, btw.</p><p>People are currently building their professional twins and uploading to them everything they’ve ever written, et cetera, and these twins are proving to be much smarter because at one go they possess at their fingertips all the information and all the knowledge that the human counterpart has, including what he may not even remember. Dial that up a notch and we could have so much data that suddenly, it’s no longer just my body of knowledge, but also knowledge of my body, my sense of humour, my peeves, my idiosyncrasies and ideologies, my nuances and habits, even the heuristics that influence my decision making over my life time. Now, that would be something I’d like to see taken on. And somebody’s doing it right now in some lab somewhere.</p><p>I’ve never really given this a second thought – until now. And it’s not just because I have replayed so many of my Uncle’s voice notes to me, or struggled with the notion that I never get to speak with one of my closest friends again until we meet in the eternal life but, I’ll admit that’s a big part of my tunnelling down this path.</p><p>However, now, almost pressingly I think the world is heading into a state of emergency where we are drastically losing the adults in the room. I’m actually concerned for Japan and then Africa in that order. As artificial intelligence heats up, it seems to be making natural stupidity a lot more glaring and ubiquitous. Unfortunately at the same time we are losing the great minds that have shaped the world across politics, art, literature, science, music, governance, culture and everything in between. The great human thinkers of our time are quickly becoming the forgotten human thinkers of the past. I would give much, for a council of elders in every national government to be constituted by the great Legacy minds of that country, as an example. What would that look like at Corporate level? What would it look like at family level? </p><p>When you watch my kind of movies, from Mortal Kombat to Marvel, A council of ancients always seems to exist in some guise or the other. The instance that I would like to reference is from Black Panther where even after his father dies, Tchalla gets to meet him again in some sort of garden of elders – although I’m pretty certain he was just high on those mushrooms they covered him with.</p><p>But, I digress. What if you didn’t need shrooms to speak with a council of elders? what if all you needed was Starlink and a Mac mini? what kind of difference would that make in a world where we are on the verge of losing the notion and substance of wisdom?</p><p>Until next time,</p><p>Spyda.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7e916eaccaca" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Innovation isn’t hard.]]></title>
            <link>https://spyda.medium.com/innovation-isnt-hard-0acc7c9def0e?source=rss-11f9cec78330------2</link>
            <guid isPermaLink="false">https://medium.com/p/0acc7c9def0e</guid>
            <dc:creator><![CDATA[Spyda]]></dc:creator>
            <pubDate>Sat, 08 Nov 2025 13:02:57 GMT</pubDate>
            <atom:updated>2025-11-08T13:02:57.083Z</atom:updated>
            <content:encoded><![CDATA[<p>The courage to make it structural is.</p><p>When Ajay Banga decided Mastercard’s real rival wasn’t Visa, but it was cash instead, he changed everything. Because that one shift wasn’t about competition.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*psZED-hzJn1xfWavrb_ylw@2x.jpeg" /></figure><p>It was about purpose.</p><p>The genius of Mastercard Labs wasn’t in its tech.It was in how it rewired a 50-year-old institution to behave like a start-up, to measure learning before revenue, and protect curiosity from the people paid to kill it.</p><p>That’s the real lesson for African organisations. Innovation isn’t a brainstorm. It’s an operating model. It’s culture.</p><p>And when you build it right, if you build it right, incumbency stops being a weight.</p><p>It becomes an advantage.</p><p>Let the records show that I am capable of writing brief posts too. Ahem.</p><p>Until next time,</p><p>Spyda.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0acc7c9def0e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Is Ai Incremental or Existential?]]></title>
            <link>https://spyda.medium.com/is-ai-incremental-or-existential-7f2545980d01?source=rss-11f9cec78330------2</link>
            <guid isPermaLink="false">https://medium.com/p/7f2545980d01</guid>
            <dc:creator><![CDATA[Spyda]]></dc:creator>
            <pubDate>Wed, 10 Sep 2025 01:14:46 GMT</pubDate>
            <atom:updated>2025-09-10T02:22:36.142Z</atom:updated>
            <content:encoded><![CDATA[<h2>Is AI Incremental or Existential?</h2><p>Is AI incremental, or existential?</p><p>Being privy to both a growing number of “AI Strategy” conversations, and a not insignificant period of study of AI disruption strategy, I’ve begun to question whether much of C-Suite leadership is asking the right questions.</p><p>We seem stuck at “How do we use AI to improve what we already do?” However, if the last month is anything to go by, the more relevant question really should be, “What if AI makes what we do irrelevant?”</p><p>While we chase marginally incremental percentages in efficiency or improvement in today’s processes, someone else is reinventing the model entirely and unlocking outcomes that are ten times greater. While we may be using AI to make customer service faster, another is removing the need for customer service altogether.</p><p>While we may be optimising supply chains, others are building supply systems that operate on completely different principles.</p><p>I understand that because this is all really new territory, the CTO/CDO/CMDO or MD is under pressure to demonstrate both strategy and results, and so these incremental figures serve great for Board reporting purposes. It’s a good strategy for “show working” and “justify your inclusion” purposes. But it’s not a winning strategy.</p><p>The real risk is playing the wrong game.</p><p>This moment requires leadership to surround themselves with minds that can reimagine from first principles, scan far beyond their own industry for threats, and treat AI as a tool for rewriting business models, not just improving them. It requires cultures that can adapt as quickly as the technology itself. It requires people who may not be cut from the organisation or perhaps even industry cloths. People who can look in and connect patterns. People for whom curiousity, innovation, disruption, tech and creativity (potentially with a hint of rebellion) are second nature.</p><p>It may also require something very new: an EDGE team or lead – a CAIO if you will – whose sole job is to sit on the fringes of the organisation, look both into, and beyond the incremental, in order to interrogate where the next wave of disruption will come from, and force the organisation to disrupt itself before it is disrupted.</p><p>This is a person or team who’s equipped and licensed to execute on the CEO/MD/Board’s mandate. A person or team whose entire job is to make their own roles irrelevant in 3–5 years.</p><p>Boards must actively begin to put one simple question on the table at their next retreat: Are we optimising an old game, or are we preparing to lead a new one?</p><p>A strategy to design a more ergonomic horse-drawn carriage is nice for the moment, but I’m afraid Mr. Ford’s building a car.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/768/1*oi_2aU8zSpUXv_FUzKyqwQ@2x.jpeg" /></figure><p>Thoughts inspired by the ever brilliant Zoe Scaman.</p><p>Until next time,</p><p>Spyda</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7f2545980d01" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>