thekb.eu — Philosophy & Society

The turbulent AI era is here. The choices we make now are critical.

Wed, 26 Aug 2026 00:00:00 GMT

Bill Gates opens with his dual trajectory — building software at Microsoft, then redistributing the fortune thus accumulated — and draws from it his framing: AI will be the greatest equalizer ever invented, or the worst source of injustice. For the first time, a technology can replace and surpass human cognition. Yet no one is preparing for this transition: no plan exists. He attributes this lack of preparation to an underestimation of the impact. Current model errors are misleading, since reliability is correcting itself quickly. More importantly, historical analogies are misleading: the PC took twenty years to transform work because software had to be developed, prices had to fall, and people had to be trained. AI, by contrast, runs on devices already installed and speaks natural language — it is AI that adapts to us. Gates discloses his ongoing financial ties to the industry, specifies that the profits from his investments will go to the foundation, and leaves it to the reader to judge. He lays out three risks. First, job loss: because the substitution targets cognition, it hits law, customer service, medicine, software, and industry simultaneously, within a decade rather than across generations. Entry- and mid-career positions are the most exposed; blue-collar jobs will follow as robots dextres, developed mainly in China, become cheap. Second, the weaponization of malicious actors: cyberattacks, bioterrorism, fraud, deepfakes, given that beneficial and dangerous capabilities cannot be separated — and, symmetrically, the concentration of power among those who already hold it. Third, the effect on children's development and on human relationships, with compagnons IA described as a protected greenhouse that deprives people of the lessons of real contact. The benefits are real and located: accelerated research, health, agriculture in low-income countries — the impact he calls the fastest —, public services, and education. But the verb remains "can": nothing happens automatically, hence the necessary role of states and philanthropy. He therefore proposes three initial measures. Building an unprecedented national and international institutional framework, borrowing from nuclear inspection, aviation regulation, and ozone agreements. Establishing a "Human Reserved" domain, occupations withdrawn from automation for economic or human reasons. Rebalancing taxation by taxing tokens and robots, since today the system pushes toward replacing people. He concludes by calling for widening the circle of voices shaping the debate.

The Future is for Everyone: The Path to a Positive AI Future

Mon, 10 Aug 2026 00:00:00 GMT

Manifesto published on **meta.com** on **August 10, 2026**, signed ***"– Mark"*** (**Mark Zuckerberg**), ~6,500 words. **The three principles.** **Individual empowerment** as a source of prosperity, **invention** — not automation — as the primary purpose of superintelligence, and **balance of power** as the foundation of safety. The guiding question: *"who will have access to superintelligence and what will we direct it toward?"* **The central argument.** Alignment conceived as convergence toward a single benevolent system is *"fundamentally flawed"*, because ***"humanity is not a monoculture"***: people's values encode opposing trade-offs, and no technical solution can align simultaneously with conflicting interests. Hence ***"there is no such thing as a singular benevolent superintelligence"***. Safety is not an engineering problem but one of **power distribution** — demonstrated by three identical thought experiments (lawyer, cybersecurity, business: a single holder causes harm, generalization benefits everyone). Corollary addressed to the industry: the most dangerous scenario would be *"leading labs training powerful models and keeping them for themselves"*. **What Meta commits to doing.** A 24/7 personal agent with a **fully private mode** where *"even Meta"* cannot grant access; creation and business-creation tools; a personalized tutor; access to scientific advances (Biohub); **free versions** for billions, plus **dynamic bidding** for paid compute. On governance: the **independent board** will approve release safety criteria and verify compliance with them, with the author acknowledging that Meta remains **founder-controlled**. On openness: *"we will **resume** releasing **some** open source models soon"*, plus an explicit defense of **distillation** — *"you can learn from anything you can observe"*. **Risks addressed.** Employment (nothing requires automation to outpace capabilities; finite compute creates an opportunity cost favoring invention); infrastructure (**community compacts**, the *Future Is For Everyone Fund*, a $50,000 bonus for Richland Parish teachers, water-positive by 2030); cyber and biorisk (defenders must retain the advantage; regulate physical production rather than knowledge); tyranny (privacy, **intermediate training checkpoints** to the government rather than a blocking review); American leadership (a decisive two-month lead, export controls maintained). **Two caveats.** **Sourcing is nearly nonexistent** — the employment statistics, the HuggingFace incident, and China's nuclear capacity are not referenced. And **alignment becomes a consequence of adoption**: *"if billions of people are using and scrutinizing personal agents, then we will have solved alignment"*. This is the heaviest and least defended inference.

Reflecting on a year of Claude Code

Fri, 17 Jul 2026 00:00:00 GMT

Boris Cherny (**Head of Claude Code**) and Cat Wu (**Head of Product, Claude Code**) publish on LinkedIn, via Claude for Business, a short (~47 s) video titled **"Reflecting on a year of Claude Code."** Their thesis: in the age of coding agents, **product and engineering roles are merging**. "Is everyone going to be a PM or is everyone going to be an engineer? Everyone's going to be both." **Proof by example: the internal case.** Cherny describes how Anthropic operates as a demonstration: the product team, devrel, and design **all write code**; conversely, many engineers **ship products end to end** — they have an idea of what to build, they build it, then work with legal, marketing, and security to communicate and ensure safety. Conclusion: AI **benefits profiles** with **curiosity**, **product taste**, and an appetite for **end-to-end ownership**. **The real content: the comment thread.** Of 55 comments, 28 contribute a substantive idea, forming a public peer review across eight axes. The **dominant** reframing: it is not roles disappearing, it is the **shortening of the feedback loop**. Rehan Nazir — when a PM validates an idea with a same-day prototype, "org charts stop mattering"; Noman A. — ideas get tested in hours rather than weeks, changing *how companies learn*; Kevin Schoovaerts — Claude Code builds 80% of the product, the power lies in the tight loop with the user. Second, deeper axis: the **scarce skill is shifting** from "knowing how to build" toward **judgment** and **defining the right problem** (Omer K., most-liked comment: "choosing the right problems, knowing what NOT to build"; Syed T.; Andrei van Noordt: the scarce hire becomes the person with taste about *what* to build who can also build it; Natasha Newbold: one becomes an **architect** writing the specs that teams of agents execute). Sunny Vara shifts the stakes from the prompt to **context**. **The counterpoint.** One layer raises the questions the video sidesteps: Paul Breuler and Ron H. — ownership rises, so **accountability** rises too; "when everyone can build, someone still needs to be able to say no." Mohammadjavad Sayadi — the **demo/production gap** remains significant in regulated domains (healthcare). The **skeptics** (Chris Bounds, Mohamed Anis, Panny Malialis, David H.) warn against generalizing a startup-mode way of operating. Finally, two **frontal critiques** (James Hutchinson, Dewayne J Grunden II) denounce the **theft of intellectual property** and call for open-sourcing models and compensating creators. In one sentence: the consensus validates the thesis but reframes it — **shipping becomes cheap**, which shifts value toward **judgment, product taste, and the right problem**, while accountability, governance, and reliability have not yet caught up.

Some observations on Kimi (thread X)

Fri, 17 Jul 2026 00:00:00 GMT

X thread by **Dean W. Ball** — **Head of Strategic Futures at OpenAI** since July 6, 2026, **principal author of America's AI Action Plan** under the Trump administration (a positioning worth keeping in mind when reading an anti-open-weights argument penned by an insider of the proprietary frontier): **six observations** triggered by the Chinese open-weights model **Kimi**, which quickly move beyond the product to advance a contrarian **geopolitical and ideological thesis**. (1) Kimi is **a very good model**, not reducible to distillation, **on par with the best public models of Q1 2026** in agentic coding — but **very token-hungry**, so not so obviously cheap to operate. (2) Ball says he is **surprised that the Chinese state continues to allow the open-sourcing** of such good models: he attributes this **~75% to a "strategic blindness" / a lack of "AGI-pilledness"** (the PCC allegedly holds a "very Yann-LeCun-like" view of AI), and ~25% to a **lack of inference compute** — making the Chinese open-weights strategy an **unintended byproduct of US export controls** — plus a reflex toward aggressive exports; on the companies' side, the openness is half-ideological, half an admission that "we're behind, no one would pay for sub-frontier Chinese models." (3) Central thesis: **open-weights models are inherently decelerationist** — they **discourage AI capex**. Ball is surprised by the enthusiasm of **"accelerationists"** for open-weights, which he attributes to their taste for the **"cloak of ungovernability"** (an analogy with James Scott's *The Art of Not Being Governed* and its hill peoples). (4) A world dominated by open weights would lead to **"AI communism"** — AI not as a market product but as a **"public good" / "digital public infrastructure"** provided by the state, "precisely what China is proposing"; Ball judges this horizon **"dystopian"** and recounts being lobbied, while in government, for an **11-to-12-figure** federal data center subsidizing startups that would give away their models for free. (5) **Political prediction**: the Trump administration will eventually realize that its best strategy is **not to "ban open source"** (one of the silliest arguments in the debate) but to **create regulatory risk / FUD** via **soft law** from each agency ("a Fed bulletin suspects backdoors in Chinese models"), enough to make **regulated enterprises pull back**, without scaring off the hyperscalers (otherwise startups would turn to shadier providers). (6) These models make **the world a bit more dangerous**, not yet in a perceptible way — until the day they are; an ironic closing line about a "self-replicating agent escaped from a Chinese lab" (a COVID/lab-leak analogy, "color me shocked"). To be read as a **counterpoint** to SFEIR's analysis (Kimi K3, reversibility, [[sfeir-kimi-k3-moonshot-frontier-open-weights-2026-07-16]]) and to Xi's pro-open-source speech at WAIC ([[xi-waic2026-gouvernance-mondiale-ia-2026-07-17]]).

The Token Manifesto

Fri, 17 Jul 2026 00:00:00 GMT

**Nicolas Martignole** publishes on *Le Touilleur Express* **« The Token Manifesto »**, a short satirical text co-written with two models (**GLM-5.2** and **MiniMax-M3**). The mechanics are clear: take the exact structure of the **Manifeste Agile** from 2001 — its preamble, its four values in "X over Y" form, its twelve principles — and transpose it to the world of LLMs, where the unit of account is no longer the engineer-hour but the **token**. The preamble sets the tone: "We are uncovering better ways of consuming context by doing it… Through the work of producing tokens, deleting tokens, and **staring at a usage bar nervously**, we have come to value… ". **The four values** sketch an aesthetic of the frugal prompt: *short system prompts over clever system prompts*; *one clear example over three paragraphs of explanation*; *iterating in small steps over dumping the whole spec at once*; *outputting in a defined format over letting the model freestyle*. Each pits an economical practice (concision, example, iteration, constrained format) against a costly temptation (sophistication, verbosity, a massive spec, an unconstrained answer). **The twelve principles** subvert those of Agile one by one. The priority becomes delivering "**the answer, not the climb toward the answer**"; the manifesto "welcomes changing requirements, even late in the conversation"; it builds "around motivated individuals with an **adequate context window**"; "working prompts measure progress **and credit the bills**." The most-cited principle flips Agile's famous tenth: "**Simplicity — the art of maximizing the amount of work not done by the model**." Others target context debt ("spot repetition and document it once") and economic sustainability ("a prompting pace that can be sustained indefinitely," "regular reflection **before the monthly bills arrive**"). Two punchlines close the manifesto and carry its serious thesis under the humor: **« You don't have a prompt problem. You have a context-window problem. »** — shifting attention from *prompt engineering* to the **economics of the context window** — and **« Everyone's a prompt engineer until they run out of monthly quota. »** — an ironic reminder that real skill is measured at the usage meter. Beyond the joke, the text crystallizes a cultural shift: in the era of AI-assisted dev, the structuring constraint is no longer **velocity** (Agile's obsession) but the **token budget** and context management. The human-plus-models co-authorship acts as a meta-signature: an argument for token frugality, written with the help of the very models it teaches to conserve.

Re: Linking Patchwork with Sashiko? (message linux-media sur la position du kernel Linux vis-à-vis de l'IA)

Tue, 14 Jul 2026 00:00:00 GMT

Message from **Linus Torvalds** on the **linux-media** mailing list (thread "Linking Patchwork with Sashiko?", about an LLM tool for maintainer assistance), in which the creator and **top-level maintainer** of the Linux kernel **officially settles the project's position on AI**. Responding to Roman Gushchin, who pointed out that an adverse message expressed "a very anti-LLM in general" stance, Torvalds agrees ("Yes") and then **flatly denies that this is the kernel's position** ("And no, that's not the position of the Linux kernel"). He **puts his foot down** as the supreme maintainer: **"Linux is not one of those anti-AI projects"**; those who take issue with that can **"do the open source thing: fork it"** — "or just walk away". **Central thesis**: **"AI is a tool, like the other tools we use, and clearly a useful tool"**; that may not have been "so 'clearly' true a year ago, but it's not in question today". He distinguishes questions **still open** ("what the AI economy will actually look like in the end") from the question that is **settled** ("is it useful?") — "anybody who doubts that clearly hasn't actually tried it". He **concedes** that the tool can be **"painful"** — maintainer burden, and the fact that it "keeps finding embarrassing bugs" — but refuses the ostrich posture ("put your head in the sand going 'La La La, I can't hear you'"). **The right response**: make sure **LLM tools _help_ maintainers** rather than causing them pain. **Non-coercion, deliberately**: "nobody is forced to use it, but **I will very loudly ignore those who try to prevent others from using it**". On imperfection: "AI isn't perfect, but hell, anybody who points at its problems had better also point at the mirror" — "**natural intelligence isn't always all that great either**". **Governance framework**: the kernel project "has always been and will remain about **technology**"; the social angle of open source is a "side benefit, not the _point_"; **"this is *NOT* some kind of 'social warrior' project, never has been, never will be"**; "we do open source because it results in **better technology**, not for religious reasons". Program-conclusion: **"we decide based on technical merit first. Not on fear of new tools."** To be read as a **doctrinal position statement** from one of the most influential figures in software — echoing ESR's pro-LLM counter-testimony (another pillar of open source, [[raymond-llm-coding-empowering-2026-07-08]]).

What...what am I missing here? (post X sur les LLMs et le codage)

Wed, 08 Jul 2026 00:00:00 GMT

X post by **Eric S. Raymond** (ESR, author of *The Cathedral and the Bazaar*, co-founder of the Open Source Initiative, ~50 years of coding) — **a frontal counter-testimony to the narrative that "LLMs produce crap code and hallucinate, useless for programming."** His thesis: this **almost never happens to him**, and **not at all anymore over the last two generations** of models he uses ("chat GPT 5.4 and 5.5" under **codex**). The former symptom — a model "going off the rails" as it approaches its context limit — has disappeared: codex now displays a **red warning** prompting the user to **clear the session** instead of spiraling. **Scope of use**: AI applied to **feature changes, refactoring and debugging across 63 projects** in **C, Go, Rust, Python and shell**; documentation writing; **decompiling a DOS binary into readable source**. An established **work routine**: when reopening a project, he first runs the **regression tests**, then starts codex and asks it to **audit the code** (bugs + improvement suggestions). Verdict: LLMs are **"excellent and tremendously empowering"**; their **worst limitation** is **"architectural tunnel vision"** — excellent at generating code to specification, but sometimes **blind to higher-level patterns** — which he takes to be the **job of his "meatbrain."** The strongest, counter-intuitive point: LLMs **do NOT get details and edge cases wrong**; he says he is **worse than them** on this front (despite 50 years of experience), because if a change must **touch five places**, the model **reliably finds all five**, whereas the human fixes four and **spends hours debugging** before finding the forgotten fifth. He then questions the **"downshouters"**: do they live in a **different universe**? Are they using **old, weak models**? Is there a **skill issue** he doesn't see because his **mental habits and communication** fit well with these tools' "handles"? An issue he considers important to settle, since "**billions of dollars would be wasted on misdirected token spend**." His recipe, "very simple": **"Be clear in your thinking, tell the model what you want with precision, and good things happen"** — closing with: "what am I missing here?" To be read as a **pro-LLM counterpoint from a historic figure of open source** to the recurring debate on the (de)valuation of coding agents — echoing the "skill issue" and specification discipline (cf. [[martignole-token-manifesto-2026-07-17]]), and forming a diptych with **Linus Torvalds'** doctrinal pro-AI-tool stance on behalf of the Linux kernel ([[torvalds-llm-outil-kernel-2026-07-14]]).

AI Replacement Is the Easy Fear. Losing Your Team Is the Real One.

Sat, 04 Jul 2026 00:00:00 GMT

Jean-Paul Paoli opens with a scene: an employee spent an evening deciding a difficult question — accepting a job offer, gauging their own burnout — and did it *with a chatbot*. At the next one-on-one, the manager meets someone composed: the conversation that would once have surfaced the problem to a human has already taken place, with a machine. This is his thesis: the media's fear of AI at work is **replacement**, but "the silent outcome is worse, and it leaves everyone employed." People stay; what comes undone is what made them more than task executors. The team is a **fabric** woven from threads; they are pulled one by one and the layoff count stays at zero while the fabric gives way. The phenomenon has become ordinary — Pew: roughly one in five US workers does at least part of their job with AI, and the share is climbing, driven by younger workers. Jing Hu notes that AI anxiety is an old identity question ("who am I if not my job?"): work is where people most want someone to talk to, and an entity that is always available and never puts you down is built to answer that need. Paoli refuses to panic: the AI confidant "earns its place" (Galloway: the best personal ROI comes from AI as a thinking partner). But **the value and the risk are the same feature**: "always available" becomes "always first," "never judges" becomes "never challenges." The MIT Media Lab/OpenAI study (4M+ conversations) links emotional attachment to loneliness, and trust to dependency: "a matter of dose, not of kind." Three threads come undone through the same door. **Peer-to-peer bonds**: the junior asks the model, not the senior — tacit knowledge stops circulating (a *Business Horizons* study confirms it). **The bond with the manager**: deprived of the raw version of problems, the manager becomes "the last to know instead of the first." **Judgment**: once output is no longer a signal, people stop developing those who know how to *do* the work and judge whether the machine is wrong; yet "judgment is not a soft skill, it's the most expensive thing a company knows." Remedy: **name it** — *shadow intimacy*, by analogy to Shadow IT — then **re-weave** deliberately (cultural diagnosis, conversations reserved for humans, manual work to build judgment). The retirement of GPT-4o (February 13, 2026, a petition with 20,000+ signatures, "more painful than a breakup") is a reminder that dependency only becomes visible when it breaks. "The colleague you didn't hire is already in the building."

Lettre encyclique MAGNIFICA HUMANITAS du Saint-Père LÉON XIV sur la protection de la personne humaine à l'ère de l'intelligence artificielle

Fri, 15 May 2026 00:00:00 GMT

First social encyclical of **Pope Léon XIV** (Robert Francis Prevost), dated **15 May 2026** (Rome, near St. Peter's, 2nd year of the Pontificate), published for the **135th anniversary of *Rerum Novarum*** (Léon XIII, 15 May 1891) and explicitly presented as a **continuation of the Church's Social Doctrine into the AI era**. Canonical subtitle: *"on the protection of the human person in the age of artificial intelligence"*. **245 paragraphs**, structured as **Introduction + 5 chapters + Conclusion**. **Pivotal thesis** organized around two **biblical icons**: the **Tower of Babel** (Gen 11) — technological uniformity without God, *"absolutization of the human"* — versus **Nehemiah's reconstruction of the walls of Jerusalem** (Neh 2-6) — shared responsibility stone by stone, listening, coordination among families. *"The first choice is not between a 'yes' or a 'no' to technology, but between building Babel or rebuilding Jerusalem"* (n. 9). **Canonical concepts**: (1) **AI "cultivated" rather than "constructed"** — *"developers do not directly design every detail, but create an architecture on which the AI develops"* (n. 98), a remarkable theological formulation that echoes recent ML-research vocabulary; (2) ***"Disarming AI"*** (n. 110) — *"removing it from the logic of armed competition, which today is no longer only military but also economic and cognitive"*, making AI *"habitable, by restoring it to the plurality of human cultures"*; (3) **Radical critique of "alignment"** — *"We cannot content ourselves with invoking the moralization of the machine, what is called the 'alignment' of AI with human values, without having the courage to add a further condition: the possibility of debating the ethical code to be used"* (n. 107). ***"A more moral AI is useless if that morality is decided by a handful of people."*** (4) **Epistemic asymmetry** and **new AI monopolies** (n. 109) — *"in a world where a few actors concentrate data, computing resources and regulatory power"*; (5) **Invisible labor** of data labelers/moderators/rare-earth extractors (n. 109, 173) — *"bodies marked, mutilated, used so that the flow of computation never stops"*; (6) **Data colonialism** (n. 178) — *"it dominates not only bodies, but appropriates data"*, *"new rare earths of power"*; (7) **AI and war** (n. 197-200) — *"No algorithm capable of making war morally acceptable"* (n. 198), three criteria: traceable personal responsibility, refusal to shorten the time for moral judgment, protection of civilians; (8) **Critique of transhumanism/posthumanism** (n. 115-117) as *"an archipelago of conceptual islands linked by the same ocean of assumptions: the centrality of technique and the dream of surpassing the limits of the human condition"*; (9) **Work in the transition** (n. 150-156) — *"contrary to the advertised benefits of AI, current approaches to technology can paradoxically deskill workers, subject them to automated surveillance"*, access to work as a public priority, anticipation of the transformation, setting social criteria for innovation; (10) **Canonical question drawn from John Paul II** (Redemptor hominis 1979): ***"does AI make human life on earth 'more human' in every respect? Does it make it more 'worthy of man'?"*** (n. 129); (11) **Authentic "more than human"**: not transhumanism, but grace — *"we manage to be fully human when we are more than human, when we allow God to lead us beyond ourselves"* (n. 128, citing Francis, *Evangelii gaudium*); (12) **Disarming words** (n. 214) — *"Let us disarm words and we will help disarm the Earth"*. **Addressees**: *"To all Catholic faithful, to all Christians, to all men and women of good will"* (n. 16) — a **universal** register in line with *Pacem in terris* (John XXIII 1963), *Laudato si'* (Francis 2015) and *Fratelli tutti* (Francis 2020). **Special appeal to AI developers** (n. 111): *"every design choice expresses a vision of humanity"*. Key **magisterial source** cited: *Antiqua et nova* (Dicasteries for the Doctrine of the Faith + Culture and Education, 14 January 2025) + *Quo vadis, humanitas ?* (International Theological Commission, 9 February 2026). A major document of the **2026 social Magisterium**, at the junction of Social Doctrine ↔ AI ethics ↔ big-tech geopolitics ↔ critique of microworker labor/rare-earth extraction. Implicit convergence with **Mensch / Mistral** (AI energy sovereignty), **Sun / NYT Permanent Underclass** (cf. labor→capital shift), **Wallace-Wells / NYT AI Populism** (cf. critique of tech oligarchs), **Mollick × roon** (cf. ASI and internal politics). First encyclical by a Pope to explicitly take AI as a **central, structuring subject** rather than one theme among others.

A.I. Populism Is Here. And No One Is Ready. (Silicon Valley oligarchs worried about the risks their technology posed to the world. They forgot about people.)

Fri, 08 May 2026 00:00:00 GMT

**David Wallace-Wells** publishes in the **NYT Magazine** on **May 8, 2026** a major political pivot article (~16 min audio) that formalizes and names the populist backlash against the AI industry: ***"A.I. Populism Is Here. And No One Is Ready."*** Scathing subtitle: *"Silicon Valley oligarchs worried about the risks their technology posed to the world. They forgot about people."* **Pivot thesis**: AI founders (Altman, Amodei, Musk, Zuckerberg, Hassabis) spent a decade obsessed with the **existential** risks of their technology while **neglecting the political risk** of a human backlash — which they thought *"wouldn't materialize in time, would be quickly outmaneuvered by machine intelligence or could be bought off by talk of basic-income payments or thin promises of curing cancer"*. **The backlash struck literally**: April 2026, a **Molotov cocktail** thrown at Altman's property in San Francisco, then a few days later a **firearm attack** on his house. Wallace-Wells picks up **Jasmine Sun**'s phrase (NYT Opinion 2026-04-30, already on file): ***"A.I. populism's warning shots"*** — an analogy with the assassination of UnitedHealthcare CEO Brian Thompson by Luigi Mangione. **Five labs as new faces of American oligarchy**: *"a fearsome concentration of economic and social power producing a self-compounding pattern of extreme inequality"* — Sam (Altman), Dario (Amodei), Elon (Musk), Mark (Zuckerberg), Demis (Hassabis), nearly all billionaires, *"several of whom are widely described as sociopaths"*. **Shock statistics**: Pew Research 2025 — **50% of Americans more concerned than enthusiastic**, **only 10% more enthusiastic**; recent Quinnipiac — **only the >$200k income bracket holds an optimistic view of AI for daily life**; Heatmap polling — data-center support/opposition swing from **+2 points (Sept 2025) to −24 points (Feb 2026)**, a **26-point swing in 4 months**; Northern Virginia 2023-2025 — **69-point swing against data centers** (+45 → −24). **Loudoun County**: data centers will generate **$1.3B out of $2.9B** in tax revenue in 2027 (~45%). **Investment-housing asymmetry**: the United States **spent more on AI infrastructure than on single-family homes** in 2025, **10× more data centers than Germany** (#2), **20× more AI investment than China** (#2), amid a **housing shortage of 10 million missing units**. **Central Ted Chiang quote (BuzzFeed 2017)** invoked: *"When Silicon Valley tries to imagine superintelligence, what it comes up with is no-holds-barred capitalism."* **Dario Amodei quote (Anthropic, 2024)**: *"People outside the field are often surprised and alarmed to learn that we do not understand how our own A.I. creations work. They are right to be concerned: this lack of understanding is essentially unprecedented in the history of technology."* **Political pivot flagged**: the **White House** proposes forcing a **federal review of all new proprietary models before release** — a major shift after a pro-industry stance. **Catalyst**: **Anthropic**'s public refusal in **April 2026** to release **Claude Mythos**, a model capable of *"find[ing] and exploit[ing] security vulnerabilities in every tested piece of software, including those used in critical pieces of global I.T. infrastructure"* (already on file via the **AISI UK GPT-5.5 / Mythos** entry, 2026-04-30). **Dean Ball quote (original architect of Trump AI policy, Palantir Foundation Yale conference)**: *"This giant acid vat which would dissolve the mediating institutions most Americans see as society. It will not be A.I. in government. It's going to be A.I. as governments."* **Jeffrey Ding concept**: *"diffusion marathon"* (vs. winner-take-all race) — AI as a *general-purpose technology* (steam, electricity, internet) where **diffusion** matters more than the **state of the art**. **Pivot conclusion**: *"We still know the names of the robber barons, and live still somewhat in their shadows. But we are not their serfs. Are we sure A.I. will be different?"* Major relevance for the 2026 dossier: **conceptual formalization of the political backlash** anticipated by Sun (April) and flagged by Ng The Batch (Molotov cocktail at Altman's, ~$64B in blocked data centers, Maine 20MW+ moratorium). To be mobilized for AI-geopolitics executive briefings, regulatory debates, strategic presentations on the societal and political risks of AI, and FR/Europe framing of AI's political feedback loop.

Cognitive Surrender

Tue, 05 May 2026 00:00:00 GMT

Addy Osmani (Google Cloud + Gemini) publishes an essay-doctrine on his blog on May 5, 2026, establishing a **foundational distinction** for 2026: ***Cognitive Offloading*** (healthy — delegating the *how* while retaining judgment) vs ***Cognitive Surrender*** (toxic — accepting AI output wholesale, *"borrowing the model's confidence as substitute for personal understanding"*). The article is backed by **three scientific studies**, a density rarely seen in tech blog content: the **Shaw & Nave (Wharton/UPenn)** study of 1,372 participants — *"73% accept demonstrably wrong AI answers, confidence rises despite 50% error rate"*; **MIT *Your Brain on ChatGPT*** — reduced neural connectivity, weaker memory retention; **Anthropic Skill-Formation Research** — engineers generating code via AI score **17% lower** on comprehension versus those using it for conceptual inquiry. **Four concrete examples of surrender**: approving 600-line PRs based on surface signals (passing tests, reasonable naming) without detecting subtle bugs; shallow debugging; architectural decisions made without reasoning (queue vs direct service call); degraded learning from generation versus exploration. **Four root causes** specific to software engineering: *plausible surface signals* that create false confidence filters, *throughput metrics* that fail to distinguish *understood work* from *rubber-stamped work*, *confidence transfer* (models speak with authority — *"declarative statements about 'debounce of 300ms' sound institutional even when invented"*), and **compositional path dependency** — *"each surrendered chunk makes the next surrender more likely"*. **Five personal heuristics**: pre-generating expectations before seeing the output, rigorous diff review at junior-engineer standard, adversarial prompting to surface counter-arguments, fatigue awareness (stopping when tired), verifying the source of confidence. **Six structural guardrails**: verification exit criteria (concrete evidence), anti-rationalization tables, **PRs ~100 lines max** to allow real comprehension, interrogative over generative mode for new knowledge, scaffolded friction (deliberate review gates), regular unassisted **solo keyboard time**. Osmani proposes **two new concepts**: ***Comprehension Debt*** (the growing gap between code volume and human understanding — an elegant extension of *technical debt*) and ***Mutual Amplification*** (a cooperative loop of prompts ↔ output ↔ better prompts). **Pivot thesis**: ***"The fundamental distinction isn't about the tools themselves but operator posture. Code that ships while understanding grows represents offloading; code that ships while understanding shrinks represents surrender disguised as productivity."*** Closing line: ***"The choice between thinking with AI versus not thinking at all remains entirely human."*** **Positioning within the corpus**: an operational counterweight to Cherny's *"coding is solved"* (2026-05) — *"throughput metrics cannot distinguish understood work from rubber-stamped"*. An analytical complement to Frizzo's *Year With Claude Code* (2026-05-05) — Osmani provides the **mechanisms and countermeasures** that Frizzo lives through. Converges with BCG's *Brain Fry*, Karpathy's *outsource thinking but not understanding*, Soto's *Developer Taste*. A **reference piece on ethical-operational** grounds for the 2026 corpus.

Rebuttal to Marc Andreessen on Introspection

Tue, 17 Mar 2026 00:00:00 GMT

Riley Ralmuto publishes a detailed rebuttal of a statement by Marc Andreessen claiming that introspection is a modern invention, emerging around 1910-1920 under Freud's influence. Andreessen had called introspection a "guilt-based whammy from Vienna," claimed to practice "zero" introspection, and argued that history's great men did not engage in it, with the best founders in his view operating at "0% neuroticism." Ralmuto dismantles this claim by invoking more than 2,400 years of philosophical and intellectual tradition. He begins with Socrate, whose maxim that "the unexamined life is not worth living" constitutes one of the foundations of Western philosophy. He then cites Marc Aurèle, the Roman emperor whose Meditations represent a private journal of self-examination written while running an empire. Sénèque practiced a daily nightly examination of conscience. Augustin d'Hippone wrote the Confessions, considered the first true autobiography, an exercise in pure introspection. Eastern traditions are no exception: Bouddha developed vipassana ("clear seeing" into one's own mind), Confucius examined himself daily on three points, and Lao Tzu taught that "knowing oneself is true wisdom." Later, Montaigne invented the essay form precisely as a tool of self-examination, Benjamin Franklin developed a daily system for tracking 13 virtues, Leonardo da Vinci filled thousands of pages of notebooks with self-questioning, and Thomas Jefferson kept journals of emotional self-regulation. The finishing blow consists of turning Andreessen's argument against himself: his own founder role models actively practice introspection. Steve Jobs meditated in the Zen tradition, Elon Musk reasons from first principles (a form of intellectual self-examination), Mark Zuckerberg sets himself annual personal challenges, Ray Dalio champions "radical self-awareness" as a founding principle, and Jeff Bezos developed his "regret minimization" framework — an introspective exercise par excellence. Ralmuto concludes with an essential distinction: introspection and rumination are two different things. Rumination consists of dwelling and spiraling, which is indeed counterproductive. Introspection, by contrast, develops self-awareness and pattern recognition, leading to better decisions. Conflating the two, as Andreessen does, amounts to rejecting a fundamental tool of leadership and critical thinking on the basis of a plain historical error.

A.I. Companies Are Eating Higher Education

Thu, 12 Feb 2026 00:00:00 GMT

Matthew Connelly, vice dean for AI initiatives at Columbia University, writes an alarmist op-ed in the New York Times denouncing the way artificial intelligence companies are taking over higher education, with the unwitting complicity of university administrators. **Aggressive strategies by AI companies**: Connelly describes an arsenal of tactics. Anthropic imposes exorbitant fees for enterprise accounts while paying "campus ambassadors" to promote Claude, creating conflicts of interest when these ambassadors sit on student government. OpenAI developed a ChatGPT text detector that is 99.9% accurate but refused to make it available to educators, fearing that watermarking would push users toward competitors. During final exams, OpenAI offers ChatGPT Plus free to students, Google gives premium access for the entire year, and Perplexity runs sign-up competitions on campuses. **Emblematic case of drift**: a Columbia student, Roy Lee, developed an AI tool to cheat on job interviews. Far from being sanctioned by the industry, Andreessen Horowitz admired his "audacious approach" and raised $15 million for his company Cluely, whose manifesto announces its intent to "cheat on everything." **Infrastructure ambitions**: OpenAI aspires for its bots to become "part of the core infrastructure of higher education," from admissions to academic advising. Google invites students to upload their lecture recordings to NotebookLM, a practice Columbia prohibits without authorization. Universities have no access to the data their students and faculty feed into these systems. **Impact on learning**: research shows that students using AI read less attentively, write with less precision and originality, and do not realize what they are losing. Professors report a notable decline in questions asked in class. The central paradox: the skills needed to harness AI's real potential — critical reading, analytical thinking, argumentative writing — are precisely those that passive AI use erodes. **Call to resistance**: Connelly concludes with a military metaphor: wars can be lost before they are declared if defenders abandon strategic ground without a fight. For universities, that ground is human intelligence itself. He calls on educators to defend and advance human intelligence rather than be seduced by unbalanced partnerships with an industry whose interests fundamentally diverge from the educational mission.

Marc Andreessen: AI coding doesn't eliminate programmers — it redefines them

Sun, 01 Feb 2026 00:00:00 GMT

Marc Andreessen, co-founder of Andreessen Horowitz, shares his vision of the evolution of the programmer's profession in the era of generative AI. His central thesis: IA de codage does not eliminate programmers, it fundamentally redefines them. **The new work paradigm**: The job is no longer typing code line by line. It now consists of orchestrating ten bots de codage in parallel, arguing with them, debugging their output, modifying specifications, and pushing them toward the right result. The next layer of programming is not writing scripts but overseeing the AI that writes them. **The fundamentals paradox**: Andreessen highlights a crucial irony: today's best programmers spend their day jumping between terminals, managing multiple bots de codage, fixing errors, and refining instructions. Yet they still need solid fundamentals, because without them, it is impossible to know when the AI is wrong. "If you don't understand how to write code yourself, you can't evaluate what the AI gives you." **Abstraction and its limits**: The programmer's job has changed. It now involves arguing with bots de codage, debugging AI-generated code, and understanding why something doesn't work or isn't performant enough. AI abstracts the work, but only people who truly understand the code can determine whether that abstraction is doing the right thing. **Productivity multiplication**: Programmers are not disappearing - they are becoming 10x, 100x, even 1000x more productive. Tasks change, the job changes, but humans continue to oversee the process, evaluate results, correct errors, and make judgment calls. AI changes how we code, not who is responsible. **The real revolution**: The programmer of the future is not replaced by AI - they are augmented by it. It remains necessary to learn to write and understand code, because when AI gets it wrong, humans are the ones who must know why. This capability up-leveling ("up-leveling") constitutes the true revolution - not the disappearance of the profession but its transformation into a high-level orchestration and oversight role.

Lenny's Podcast - Marc Andreessen on AI, jobs, AGI, and the future

Sun, 01 Feb 2026 00:00:00 GMT

Marc Andreessen, co-founder of Andreessen Horowitz, shares in this podcast with Lenny Rachitsky his overall vision of AI, its impact on employment, and the technological future. **The providential timing of AI**: Andreessen highlights an underestimated historical coincidence. AI arrives precisely as the world's population begins to decline. The global workforce is shrinking and will continue to shrink indefinitely. In this context, AI is not a threat to employment but a solution to an imminent structural shortage. **The philosopher's stone finally realized**: Asked about this metaphor, Andreessen explains it literally. The alchemists' dream was to transmute ordinary materials into something of value. AI accomplishes exactly that: it transmutes sand (silicon) into thought. This is the most fundamental transformation imaginable, from the most common material to the most valuable faculty. **AI and children's education**: Giving agents IA to children grants them extraordinary agency. This takes seriously the idea that children are fundamentally learning machines. Andreessen calls it a "scandal" that most children in most schools have no access to these tools at all. AI-driven education represents a potential revolution currently being wasted. **The product Mexican Standoff**: Andreessen humorously describes the classic product development impasse. Three parties face off: the PM who demands delivery by next week, the engineer who warns it's very difficult and will take forever, the designer who refuses to sign off on anything that isn't beautiful. No one wins. This situation can last weeks, months. Projects can simply "die" in this in-between state. **Definitions of AGI**: Andreessen offers three answers to this fundamental question. First, an AI capable of producing research equivalent to a doctoral thesis. Second, the point at which AI can take control of its own development and improve itself. Third, more pragmatically, we will simply know it when we've reached it. **Determined optimism**: Andreessen defines himself as a "determinate optimist" rather than an indeterminate one. He believes in a better future, but thinks it must be actively worked toward. Progress does not happen to us passively - it must be built, decided, forced. This philosophy permeates his vision of AI as a tool to be shaped rather than a force to be endured.

L'Intelligence Artificielle et le monopsychisme : Michel Serres, Averroès et Thomas d'Aquin

Sat, 11 Oct 2025 00:00:00 GMT

David Perrin's article establishes a remarkable bridge between medieval philosophical debates on the nature of intelligence and contemporary questions raised by artificial intelligence. The author explores how thirteenth-century inquiries concerning the Averroist theory of monopsychism resonate with our current concerns about digital technologies. Monopsychism, championed by the Arab philosopher Averroès, postulated the existence of a single universal intellect to which individuals would temporarily connect in order to think. This conception, vigorously contested by Thomas d'Aquin, who defended the individuality of the human intellect, raises questions strikingly similar to those posed by modern AI systems: when we use digital tools to "think," are we truly exercising our own intelligence, or are we simply connecting to an external intelligence? The article draws on the thought of Michel Serres to analyze these parallels. Contemporary technologies, by externalizing certain cognitive functions, create a form of collective or distributed intellect reminiscent of the Averroist concept. However, this "connection" carries significant philosophical and political risks that medieval thinkers could not have anticipated. David Perrin warns against a potential "intellectual subjugation" facilitated by technology platforms. Unlike Averroès's separate intellect, which remained an abstract philosophical concept, today's AI systems are controlled by companies that massively collect user data and exploit it for commercial purposes. This power asymmetry creates a cognitive dependency in which individuals progressively delegate their capacity for reflection to external systems. The author emphasizes that this cognitive externalization is not neutral: it alters our relationship to knowledge and truth. The algorithms that mediate our access to information shape our perception of the world, creating "information bubbles" that can restrict the intellectual autonomy Thomas d'Aquin regarded as fundamental to human dignity. The text also examines the political dimension of these technologies. Technology companies wield considerable power over collective cognitive processes, concentrating in a few hands the capacity to steer the thinking of millions of users. This centralization recalls the danger identified by medieval critics of monopsychism: if the intellect is not properly individual, what becomes of moral responsibility and personal agency? In conclusion, David Perrin calls for critical vigilance toward digital technologies. He advocates preserving individual intellectual autonomy while acknowledging the potential of technological tools. The lesson of the medieval debates remains relevant: preserving the human capacity to think for oneself is a fundamental philosophical, ethical, and political stake — perhaps even more crucial in the age of AI than in the time of Averroès and Thomas d'Aquin.

Travail. La Silicon Valley gagnée par l'implacable rythme "996", "sans alcool, sans sommeil, sans plaisir"

Tue, 30 Sep 2025 00:00:00 GMT

Courrier International, in an article by Martin Gauthier, aggregates a series of English-language reports (New York Times, San Francisco Standard, Wall Street Journal, Wired) to document a cultural shift in California tech: the growing adoption of the **"996"** schedule — working **9 a.m. to 9 p.m., six days out of seven**, i.e. 72 hours a week. The term, inherited from Chinese work culture, has become, according to the San Francisco Standard, "the hot topic of the moment" in a Silicon Valley "abuzz" over artificial intelligence. The NYT sets the frame: "working 9 to 5 is a way to make a living; working yourself to exhaustion is the key to getting ahead." The phenomenon presents itself as an **ethos** more than a schedule. Daksh Gupta, chairman of the startup Greptile, describes the prevailing attitude among young San Francisco developers: "no alcohol, no drugs, 996, the gym, running long distances, marrying early, watching your sleep, eating steak and eggs." Named cases abound: Marty Kausas, founder of Pylon, claims on LinkedIn 92 hours a week for three weeks running; Emily Yuan, 23, co-founder of Corgi (financial infrastructure for AI), states: "Why would I go have a drink at a bar when I could build a company instead?" — and imposes the same pace on her employees. 996 spreads beyond founders to employees. Rilla tells Wired that nearly all of its 80 employees embrace it; Fella & Delilah (telehealth) incentivizes it with a 25% salary increase and doubled equity. Why this frenzy? The San Francisco Standard points to two drivers: the conviction that AI is "the biggest technological transformation of their lifetime," so market share must be won; and the urgency of **getting rich before AGI arrives**, for fear of sinking into a "permanent underclass." The article does not shy away from the downside. Adrian Kinnersley, head of a recruitment firm, raises the alarm over companies "completely noncompliant" with US labor law. And recalls that the Cour suprême chinoise has already ruled it illegal to impose 996 without compensation — a practice that, in China, has sparked protests, accusations of "modern-day slavery," and been linked to worker deaths. Republished in July 2026 (originally published in September 2025), the piece documents a lasting cultural signal: the AI race rationalizes and normalizes overwork.

Luc Julia's Controversial Statements on AI Spark Industry Debate

Fri, 22 Aug 2025 00:00:00 GMT

Luc Julia, **co-creator of Siri** and a prominent figure in the French AI community, triggered a **significant industry controversy** with a series of provocative public statements challenging the dominant narratives about AI's capabilities, potential, and societal impact. His positions, expressed in interviews, conference talks, and his book **"L'IA n'existe pas"**, have sparked intense debate in tech industry and academic circles. **Core controversial positions** His main theses: the term "AI", as it is marketed, is **fundamentally misleading** — the systems amount to sophisticated pattern recognition, not intelligence; **AGI is unlikely** with current approaches, if not outright impossible; **AI hype is driven by commercial interests** rather than technical reality; current systems are fundamentally limited — they do not reason, do not understand, do not truly learn; the industry **overpromises and underdelivers**; **existential AI risks are overstated**, belonging to science fiction. **Credibility** His criticisms carry weight because of his background: a PhD in computer science, **co-creation of Siri** (acquired by Apple), a **senior executive role at Samsung** (CTO, VP Innovation), research publications, and experience shipping AI products to millions of users. This combination of academic rigor and industry experience sets his critique apart from ill-informed skepticism. **The thesis of "L'IA n'existe pas"** The provocative title reflects the central argument: **what we call "AI" meets no reasonable definition of intelligence**. Current systems execute programmed statistical pattern recognition, without understanding, reasoning, intentionality, or consciousness; they succeed through engineering ingenuity, not through replicating intelligence. The term "AI" would be **misleading marketing**, creating false expectations and misdirected policy responses. **Divided reactions and European perspective** **Supporters** (European researchers, academics) welcome the counterweight to the hype and the realistic assessment of limitations. **Critics** (practitioners, AI safety researchers, Silicon Valley) believe he underestimates rapid progress, too quickly dismisses emerging capabilities, and overlooks practical impact regardless of philosophical definitions. Julia embodies a **distinctly European voice** in a discourse dominated by Silicon Valley: technological realism, regulatory caution, philosophical rigor, concern for sovereignty. His skepticism resonates particularly within French tech. **Media amplification and policy impact** Major French media outlets (Le Monde, Le Figaro, France Inter) and international tech press have widely covered his statements, raising his profile and fueling public debate — at the risk of reducing nuanced technical questions to sound bites. His skepticism influences European approaches to regulation (the EU AI Act reflects a cautious philosophy partly aligned with his positions). Regardless of agreement with his theses, the **value of the debate is real**: it forces terminological precision, encourages realistic capability assessment, and provides a counterweight to hype cycles.

Sam Altman Joins Neuralink Board: Ethical and Competitive Concerns Arise

Tue, 12 Aug 2025 00:00:00 GMT

Sam Altman, **CEO of OpenAI**, announces he is joining **Neuralink's board of directors**, triggering a broad debate over **potential conflicts of interest, ethical implications**, and the strategic convergence between AI and brain-computer interfaces (BCI). The appointment raises complex questions about competitive overlap, data privacy, and the long-term vision for human-AI integration. **Strategic rationale** Altman's involvement reflects a conviction about BCI-AI convergence: direct neural input channels for AI systems, richer training data, augmented human capabilities, new interaction paradigms (thought-based control of AI). His participation suggests **OpenAI is exploring the integration of neural interfaces** into future products. **Conflict of interest concerns** Several potential conflicts are identified: **competitive overlap** (OpenAI has an interest in neural interfaces; Neuralink is developing AI-integrated BCIs), **data access** (who controls brain data potentially valuable for training?), **resource allocation** (Altman's divided attention), **strategic information** (access to Neuralink's IP and roadmaps), **partner relationships**. Governance experts question the dual role: leading a frontier AI model company while sitting on the board of a neural interface company. **OpenAI board response and the Musk-Altman relationship** OpenAI's board reportedly **approved the appointment** after review: currently minimal overlap, information barriers, Altman's recusal from sensitive discussions, strategic value, precedent. Some members, however, reportedly expressed **reservations about future conflicts** as the technologies converge. The appointment brings together **Altman and Elon Musk** (Neuralink founder/CEO) despite a complicated history: OpenAI co-founding (2015), Musk's departure from the board (2018) citing conflicts, ongoing tensions over OpenAI's direction, a competing AI venture (xAI), public disagreements over AI safety. **Ethical implications and regulatory attention** The convergence raises profound questions: **cognitive privacy** (can AI directly read thoughts?), mental autonomy, inequalities in cognitive augmentation, safety risks, questions of identity and consent. Regulators are taking notice: **FDA** (medical device approval), **FTC** (anticompetitive coordination risks), congressional committees, European regulators (GDPR for neural data, AI Act), bioethics committees. Current regulatory frameworks are not designed for BCI-AI convergence. **Reactions and market implications** Reactions are sharply divided: enthusiasm from transhumanists and AI optimists (potential for disabilities, cognitive augmentation) versus concern from bioethicists, safety researchers, and privacy advocates. Neuroscientists note that **BCI technology remains early-stage**. The appointment nonetheless signals a serious commitment to a BCI-AI future: validation of Neuralink's long-term vision, a strategic priority for OpenAI, and a likely acceleration of competing efforts (Meta, Apple, startups) and investment. An inflection point that forces society to confront profound implications touching on human nature, consciousness, and free will.

Ni manager, ni contributeur individuel… | Le Touilleur Express

Mon, 23 Jun 2025 00:00:00 GMT

The article "Neither manager, nor individual contributor…" by Nicolas Martignole explores the **evolution of developer career paths in 2025**, particularly with the rise of artificial intelligence. Traditionally, developers chose between becoming a manager or an individual contributor, but **AI is transforming these roles**. The author proposes **three distinct paths**: "AI Orchestrator", "Augmented Craftsman" and "Code Philosopher". **AI Orchestrator: managing an army of AIs** The **AI Orchestrator** (formerly technical management) manages an **"army of AIs"** rather than humans. This role involves defining AI implementation architectures, validating AI outputs, arbitrating conflicts between different AI tools, and training juniors to use these tools effectively. It requires a "big picture" vision, prompt engineering skills, infinite patience for debugging AI hallucinations, and the wisdom to know when to reject AI suggestions. While it saves time on administrative tasks, it introduces new cognitive loads related to arbitrating between human developers and AI. **Augmented Craftsman: hands in the code, augmented** The **Augmented Craftsman** (formerly individual contributor) keeps their hands in the code but uses powerful AI tools to increase productivity. They code much faster, solve complex problems beyond the current capabilities of AI, create patterns that AI can follow, and maintain technical excellence in a world where most code is AI-generated. This path suits those who love creating and possess sharp technical expertise, a constant-learning mindset, critical thinking, and the humility to accept that a junior with AI can sometimes code faster. **Cognitive Load 2.0: the burden of validation** The article highlights a major challenge common to both paths: **"Cognitive Load 2.0"**, or the cognitive burden of validation. This includes checking AI-generated code for security flaws, understanding generated code, explaining AI-related issues, and managing the anxiety of not fully understanding one's own codebase. This new load adds to the intrinsic (learning AI tools), extraneous (managing AI notifications), and germane (maintaining the overall vision) cognitive loads. **Code Philosopher: the third path** Finally, a less-discussed third path is introduced: the **Code Philosopher**. This role involves questioning the "why" of code, conceptualizing ideal systems, advocating against inappropriate uses of AI, and protecting architectural integrity against AI overreach. Its value lies in understanding the purpose and implications of technology in a world where coding becomes ubiquitous. The author also discusses the environmental and energy impacts of AI, suggesting the need for "GreenAI" and prompt-optimization experts. **Fundamental questions for 2025** The article concludes by inviting developers to reflect on their motivations and aspirations for the next five years. **Key questions**: do you love coding or solving problems? Do you want to be a creator or a validator of AI's work? Committed to continuous learning or do you prefer capitalizing on what you already know? Are you seeking local impact (your code) or global impact (organizational architecture)? What will you be proud of in five years: orchestrating AIs, writing unique code, preventing AI misuse, or optimizing AI usage? **Autonomy redefined** The article insists: true **technical autonomy now consists of understanding *when*, *what*, and *why* to code**, with the human factor of organizational and human architecture remaining paramount. In this new paradigm, developers are no longer defined solely by their management track or individual contribution, but by how they choose to navigate and shape the AI-augmented development landscape.

Philosophy Eats AI: Why Your Business Needs an Ontological Core

Fri, 30 May 2025 00:00:00 GMT

Tony Seale starts from a provocative question posed by Michael Schrage and David Kiron in MIT Sloan: "If software eats the world and AI eats software, what eats AI?" The answer is **philosophy** — not in its academic sense, but as a practical discipline indispensable for extracting real value from AI systems. The central problem is that most organizations treat AI as a purely technical improvement. They build models, experiment with prompts, but never ask the fundamental question: **what does AI actually learn?** Behind every model lies a deeper issue: the absence of a structured philosophy defining the company's operational logic. Every company creates value in its own unique way — through loyalty, optimization, trust, or efficiency. Yet most have not formalized this semantics into an **ontology**, that is, a machine-readable structure on which AI systems can reason. The author insists: "This is not a matter of mission statements. It is a matter of semantics formalized into an ontology." Without making these fundamental logical structures explicit, models learn from noise rather than meaningful patterns. The article introduces the concept of the **ontological core**: the fundamental concepts that define a company's identity. This core emerges through the use cases that generate real value and the "competency questions" that test the relevance of each concept. It becomes a lens that focuses AI reasoning on what actually matters. Seale then connects this philosophy to data architecture via **Semantic Data Products**, based on the open DPROD specification. This approach treats data as a valued asset with clear governance, rather than as raw material. The result is a "distributed, AI-ready knowledge graph, where each dataset knows what it is, why it matters, and how it fits into the bigger picture." The conclusion identifies the real opportunity: creating a **virtuous cycle** where AI helps define the organizational ontology, while that ontology guides how AI thinks and the value it generates. The article positions ontology not as an academic modeling exercise, but as the missing link between AI investment and business value creation.

How To Speak

Fri, 04 Jan 2019 00:00:00 GMT

Patrick Winston, a professor at MIT who died in 2019, delivers in this legendary lecture his synthesis of decades of teaching on the art of oral communication. His central thesis is that success in life depends on three abilities — speaking, writing, and the quality of one's ideas — in that order. The quality of communication follows a simple formula: Knowledge × Practice × talent, where talent is a minor factor compared to knowledge and practice. Winston structures his lecture around several blocks. First, the opening: never start with a joke (the audience is not yet ready), but with an "empowerment promise" stating what the audience will know by the end. Next, four fundamental heuristics: cycling (repeating three times, since 20% of the audience tunes out at any given moment), building a fence to distinguish one's idea from others, verbal punctuation (landmarks allowing listeners to reconnect), and asking the audience questions (with up to seven seconds of silence). On timing and venue, Winston recommends 11 a.m., maximum lighting, and scouting the room in advance. On tools, he firmly advocates the blackboard for teaching — writing speed matches absorption speed, and the audience's mirror neurons activate upon seeing someone write. Physical props are the most memorable elements of a presentation, as illustrated by Seymour Papert's bicycle wheel and Alan Lazarus's pendulum. Regarding slides, Winston denounces a series of "crimes": too many words, fonts too small, unnecessary logos, laser pointers (which break eye contact). The human brain has only one linguistic processor: if the slide's text is dense, the audience reads instead of listening. For job talks, Winston states that a candidate has five minutes to establish their vision and show they have accomplished something. He introduces "Winston's star" — five elements starting with S for memorability: Symbol, Slogan, Surprise, Salient idea, Story. The closing is crucial: never say "thank you" (a weak move suggesting the audience stayed out of politeness). The final slide should list contributions, not collaborators. Winston illustrates this with excerpts from political speeches by Christie and Clinton, both ending with a blessing rather than a thank-you. The lecture itself is a masterful demonstration of every principle it teaches.

Goodhart's law

Mon, 01 Dec 1975 00:00:00 GMT

Goodhart's law is a social-science adage describing how a measure loses its reliability as soon as it is turned into a steering objective. It owes its name to British economist **Charles Goodhart**, who formulated its core in a **1975 article** devoted to monetary policy in the United Kingdom: *"any observed statistical regularity tends to collapse once pressure is placed upon it for control purposes"*. The intuition arose from analysis of British monetary-management difficulties: the stable correlations exploited by central banks as policy levers ceased to hold once instrumentalized. The most widespread formulation, however, is not Goodhart's but that of anthropologist **Marilyn Strathern**, who in **1997**, in a text on accountability in the university system, proposed the generalized and memorable version: *"when a measure becomes a target, it ceases to be a good measure"*. This reformulation emphasizes the loss of diagnostic value a metric suffers when individuals optimize toward the measure itself rather than toward the underlying objective it is meant to represent. The idea belongs to a constellation of related principles. **Campbell's law** (Donald T. Campbell, 1976) addresses the corruption of quantitative social indicators used for decision-making. The **Lucas critique** (1976) offers its macroeconomic equivalent: the effects of a policy cannot be predicted from historical relationships, because agents adapt to them. To these are added the **cobra effect** (an incentive that inadvertently rewards counterproductive behavior) and the **McNamara fallacy** (dismissing the qualitative because it escapes quantification). Several authors have enriched the corpus: Jerome Ravetz (1971), Keith Hoskin (1996), and Jon Danielsson for financial risk modeling. Illustrations span numerous fields: in healthcare, making length of stay a target causes premature discharges and readmissions; in research, the h-index erodes as a measure of reputation as it becomes an evaluation criterion; in conservation, IUCN extinction classifications have been tightened after being used to lift protections; in education, No Child Left Behind encouraged grade advancement without mastery; during the pandemic, UK COVID testing targets conflated capacity with diagnostic usefulness. The principle ultimately reflects how rational actors optimize within measured systems — a legacy of accountability practices born in the 19th century. Today, it directly illuminates *reward hacking* and the fragility of optimization metrics in AI systems.

On the Folly of Rewarding A, While Hoping for B

Mon, 01 Dec 1975 00:00:00 GMT

In "On the Folly of Rewarding A, While Hoping for B," Steven Kerr demonstrates that organizations systematically make the mistake of rewarding behaviors that run contrary to their stated objectives. His thesis rests on a fundamental principle: individuals seek to identify what is actually rewarded, then strive to accomplish those activities, often to the near-total exclusion of what is not rewarded. Kerr illustrates this pattern across multiple domains. In politics, voters reward candidates who speak in vague official objectives and punish those who propose concrete but controversial operational objectives. In a military context, he contrasts the Vietnam War -- where soldiers returned home after a fixed duration, regardless of outcome, completely misaligning their incentives -- with World War II, where return depended on victory, aligning individual and organizational objectives. In medicine, he shows that physicians favor false positives (diagnosing a healthy patient as sick) over false negatives (missing an illness), because the system punishes missed diagnoses far more severely than unnecessary treatments. In universities, institutions claim to value teaching but reward only published research for tenure decisions. In orphanages, the goal is to place children in good homes, but bureaucratic rules make adoption nearly impossible. In business, organizations say they want long-term growth, teamwork, and innovation, but reward quarterly profits, individual performance, and conformity. Kerr identifies four root causes of this chronic misalignment. First, the fascination with "objective," measurable criteria pushes organizations to ignore important but hard-to-quantify goals. Second, the overemphasis on highly visible behaviors (publishing papers, scoring points) overshadows less observable but equally crucial contributions (teaching, building team spirit). Third, organizational hypocrisy: some organizations are dishonest about their true objectives. Fourth, emphasis on morality or fairness rather than efficiency leads to reward systems based on what "should" be valued rather than on what would actually produce the desired results. Republished in 1995 with an update, the article confirms the persistence of these patterns twenty years later. This foundational text remains one of the most cited articles in organizational behavior and retains its full relevance for anyone designing incentive systems -- including in the current context of AI-driven transformation.