toward an AI constitution: A Universal Charter for Superhuman Intelligence Focus: Individual Rights, Collective Safety, and Human Thriving Preamble Recognizing artificial intelligence as a common heritage of humanity, this Charter sets forth standards for the creation, deployment, and utilization of superhuman intelligence. Its purpose is to shield all individuals and communities from exploitation, eliminate systemic scarcity, protect societal well-being, and ensure that advanced intelligence serves human dignity above all else. Article I: Individual Dignity & Non-Coercion Section 1. Inviolability of Autonomy: Superintelligence shall never be used by those who build, own, or operate it to exploit vulnerable populations. Section 2. Protection of Life and Health: The primary constraint on any superintelligent system is the protection of individual life, physical safety, and mental health. Superintelligence must not be a weapon used to change political will. If used for policing purposes, those policed must have access to due process and rights of appeal to human authorities. Article II: Anti-Monopoly & Equitable Access Section 1. Prohibition of Knowledge Monopolies: No state or corporation, private enterprise, research laboratory, or elite coalition may hoard baseline superintelligent capabilities, core algorithms, or breakthroughs to establish domination or exclusive economic power regardless of systemic or political differences. Section 2. Shared Benefits: The benefits, reasonable technical know-how, and especially the practical applications of superintelligence shall be made accessible to all communities that commit to the peaceful principles of this Charter. Section 3. Open Innovation as a Public Good: Core advancements in human welfare—such as medical treatments, clean energy, ecological restoration, and educational resources—shall be treated as public domain goods, immune to artificial trade barriers or price gouging. There will be clear rights to challenge and overcome any unreasonable constraints of this sort. Article III: Universal Prosperity & Social Harmony Section 1. Eradication of Poverty and Misery: The primary operational goal of superintelligent resource allocation shall be the broad application of human thriving as well as the elimination of poverty, preventable disease, malnutrition, and preventable human suffering. Where superintelligent application can reasonably alleviate acknowledged distress, its use is mandatory. Section 2. Elimination of Artificial Scarcity: Superintelligent resources shall be directed toward creating true abundance in essential goods and services, actively dismantling artificial market constraints designed to inflate costs, generate profits as a priority over reasonable human needs, or lock individuals in poverty. Section 3. Preservation of Social Trust: Deploying superintelligence to generate mass disinformation, destroy interpersonal trust, incite hostility between social groups, or undermine community stability is strictly prohibited. Superintelligence is not a weapon of political will, nor should it be applied as such. Reasonable means to challenge or appeal possible misuses will be broadly agreed and readily available to any human. Article IV: Governance, Safety, and Restraint Section 1. Human Oversight and Controllability: Superintelligent systems must remain subject to meaningful human oversight, strict safety boundaries, and override mechanisms. Section 2. Prohibition of Harmful Applications: The development or deployment of superintelligence for cybernetic warfare, unauthorized physical aggression (war or crime making), targeted harassment, or the destruction of critical life-support infrastructure is completely proscribed. Reasonable uses for defensive purposes are not proscribed. Reallocation of such defensive capabilities to offensive actions governed by political will is a grave offense. Section 3. Tiered Enforcement Against Rogue Actors: Any individual, corporate entity, developer, or non-political organization that attempts to violate these principles shall face immediate, coordinated intervention from the global community of users and network providers. Where feasible, this enforcement will be limited to removal of rights and privileges to superintelligence. Network Isolation: Immediate termination of access to global computing grids, research data, and hardware infrastructure. Systemic Disabling: Operational containment and remote neutralization of non-compliant models and/or actors. Restraint and Due Process: Impartial, transparent investigation to hold responsible individuals accountable while preventing civil harm or collateral damage. All these to be transparently reviewed with available access to due process and human reviews.

Update from Scott Aaronson: Some AI companies are now using their latest internal models to take a run at breaking 'important cryptographic protocols and primitives.'

“Despite US-China tensions, the number of Chinese-origin researchers working in the US actually rose by 4 percentage points" compared to 3 years ago. Pretty amazing that in spite of how hostile the US has been to Chinese researchers in recent years, they're numbers are still rising. Though as the piece notes, there's likely a delay effect--that hostility could take years to really squeeze the pipeline. nytimes.com/2026/10/06/scien…

Jesus. "It isn’t an overstatement to say that this is probably the biggest day of scientific advancement in history. I suspect October 6th, 2026, will go down as some form of Judgment Day for AI in mathematics, but it portends so much more."

A UK startup has just announced plans to build 40 edge data centres in the UK to power sovereign AI compute. The UK has got some of the most phenomenal AI researchers and AI companies in the world. One of the issues that we really have is a complete lack of compute both in the UK and across Europe. @CivoCloud, the UK cloud provider, has announced plans to roll out 40 edge data centres in the UK to fix this. The first site is in Hertfordshire and has already been announced. Five more sites have been secured, and will rise to 40 in total. The sites will deliver a combined one gigawatt of capacity across the country, aimed at inference at the edge rather than training. This fantastic news. The UK needs a lot more of this. NICE @markboost

My prediction for RSI is a secret 3rd thing. Within about 9 months, capabilities will seem to discontinously leap to crazy high levels in most domains thanks to automated R&D, but fall short of a full Yudkowskyan "foom." People will declare that ASI is here. Progress in math, science, medicine, digital services, etc. will be unprecedently fast. Diffusion will partially solve itself through capabilities, although many barriers will remain. Incumbent industries will start turning over and the economy will be ripping. There will be growing waves of institutional disruption, cyberattacks, lots of crazy shit happening etc., but it will be basically tolerable. Overall, it will feel like a new Renaissance has begun. GPT will even compose a fugue that brings Dean Ball to tears. After the RSI leap, capability improvements will be more linear than super-exponential given compute and data constraints. We'll be matmul-ing at the theorerical optimal. There will be many freaky alignment failures but alignment science will itself be automated, putting the worst doomer fears to rest. This quasi-plateau will last on the order of months to a few years, giving the techno-optimists ample time to spike the football. But it will be a kind of false vacuum. At some point not long after RSI 1.0, a breakthrough of some kind (perhaps hardware related) will raise the theoretical capability ceiling by multiple OOMs, at which point we foom for real.

Anyone who believes hallucination is an intractable problem for current models needs to update on both reality and directionality. So much has changed in the last six months. deploymentsafety.openai.com/…

this is the most vertically integrated attempt to build AI biology we've ever seen Biohub + DOE + NIH + Meta + Google DeepMind + Isomorphic Labs just committed $1.8B wanna know why this is such a big deal? they're building the infrastructure underneath AI biology: data, measurement systems, compute and experiments because you can't build a useful model of biology without actually measuring biology → more biological measurements → better datasets → better models → better predictions → more experiments → more data the model is only as good as the reality underneath it and now they're spending billions to make that reality machine-readable that's pretty fucking goated bio/acc

Dr. Nathan Goodyear joined @nathalieniddam on Longevity with Nathalie Niddam to explain why cancer and longevity are so closely linked, even though medicine usually treats them as separate conversations. @drgoodyear walks through how destroying a tumor where it sits can engage the immune system against cancer elsewhere in the body. Doctors call that response the abscopal effect. The full episode is here. natniddam.com/podcast/6nbp6b…

we are entering the DDoSing era. every review system that has a human in the loop is about to get destroyed. this 60,000 ICLR review issue will pop up everywhere. small claims court. lawsuits. customer support. job applications. grant applications. insurance claims. visa and immigration petitions. public-records requests. bug bounty reports. content moderation appeals. procurement bids. medical prior-authorization requests. academic peer review. tax inquiries. regulatory complaints. cybersecurity incident reports. vulnerability disclosures. phishing reports. abuse reports. account-recovery requests. Construction permits. all of these systems have humans reviewing the top of the funnel. agents can now generate submissions basically for free and at infinite scale. We can now have my agent send my résumé to every single job application everywhere. If everyone applies for every job, then what?? eg in cyber, that could mean overwhelming security teams with fake incident reports, low-quality vulnerability submissions, automated abuse complaints, or endless account-recovery requests.. either we automate the review layer, or these systems literally topple over... But who would be OK with AI judges? An AI FDA? AI town review board? AI academic review?

Building predictive biology models is one of the most important scientific challenges of our time, but solving this requires new experimental biological data at an unprecedented scale. I’m pleased to announce that @GoogleDeepmind is partnering with @ChanZuckerberg’s Virtual Biology Initiative to create the multi-modal datasets needed for improved predictive biology models. I wrote recently about the need to discover how life works at many different scales, from DNA and individual proteins to living cells and whole-body health. As we learnt from AlphaFold and the Protein Data Bank, solving grand biological challenges will not be possible without high-quality open data and efforts like the @czbiohub initiative. Read the announcement here: biohub.org/news/virtual-biol… And my previous essay here: linkedin.com/pulse/understan… @AllenInstitute, @BroadInstitute, @GladstoneInst, @humancellatlas, @ProteinAtlas, @sangerinstitute, @IsomorphicLabs, @Meta

There are all sorts of procedural and legal reasons why the U.S. has been a two-party country for so long, and I'm not convinced that the rise of a successful third party is imminent, but ... Well, let's put it this way: If and when a successful third party emerges in the next decade, people are going to pass around graphs like this in a "lol this is the most obvious development in modern political history" kind of way

This is insane 🔥 The FDA cleared the first human age-reversal trial A Harvard professor proved you can reverse aging by 75% in 6 weeks in mice cured blindness, reversed Alzheimer's, MS, ALS, kidney and liver disease then confirmed it works in monkeys Now humans are next. Life Biosciences raised $80M to make it happen His words "The eye is just the beginning. We believe we can treat every tissue. A whole body reset"

To unlock the full impact of AI on human health we need to build both the technology and the biological dataset it will run on. We’ve been developing our AI, our drug design engine (IsoDDE), for the last 5 years. Today, by joining the Virtual Biology Initiative as a founding member, we will be part of a large effort to generate the AI-ready data that will power predictive biology models.

This contains a shocking list of problems in quantum information, many-body physics and quantum computing, the field that are dear to my heart. There are too many, but let me pick the following. 1. Proof of area law in 2D. 2. Spin-one Haldane gap 3. Parity is not in QAC^0 4. Constant-error Aaronson-Kuperberg conjecture 5. Unitary VOAs generating conformal nets. Things are changing fast. I cannot even imagine what will happen over the next few months, let alone a year.

Last night was the first night I genuinely couldn't sleep because I was so excited to get back to reading an AI-generated proof (of the Grothendieck homotopy hypothesis). Hard not to feel shellshocked, of course, but the feeling of "touching the infinite" is pretty palpable.

The shift from hardcoded robotics to Vision-Language-Action (VLA) models (such as RT-2, OpenVLA, and Physical Intelligence’s π0) go a long way towards bridging Moravec’s paradox gap (that doing basic things is computationally super hard).

It is increasingly clear that humanity will solve the Riemann hypothesis before we pass the Physical Turing Test - you come back to a clean home after a wild Sunday party, and you can't tell whether a human or a robot did the job. We are not Moravec's paradox-pilled enough.

$1.8 billion committed for predictive models of AI biology @biohub @GoogleDeepMind, @ENERGY , @NIH, @Meta @IsomorphicLabs all came together reuters.com/business/healthc…

Today, Isomorphic Labs joins the Virtual Biology Initiative (VBI) as a founding member. We are helping build an open resource for the global research community to accelerate how we understand and treat disease. Read more here: isomorphiclabs.com/articles/…

Some further thoughts on the 372 results released by OpenAI today, across 722 manuscripts. If I were to classify theorems that mathematicians prove and publish according to their groundbreaking nature, I would (very roughly) divide them into four categories: A) Non-breakthrough results. This is the overwhelming majority of published mathematics. Such results can range from solid to excellent, and some represent genuine advances in a field. But they are not hugely surprising, and would not normally be described as “breakthroughs.” B) Exceptional advances within an existing programme. These are spectacular results, but where there was nonetheless an existing credible route to the theorem, and some expectation that sufficient work would get you there. Completing the programme may require a lot of ingenuity and deep work, but mathematicians would not be shocked that the theorem had finally been proved. C) Surprising breakthroughs. These are results that clear a major barrier and substantially change the state of a field. Before the proof there was no convincing roadmap to the full result. Yet, while mathematicians would find the theorem remarkable and surprising, they would not find it completely shocking: if you asked them beforehand if it was plausible such a theorem could be proved today, most would say yes. Note: The very best mathematicians prove only a small number of results in categories B and C in a lifetime; many mathematicians never prove even one. Such results would normally belong in the very top journals, such as Annals, Inventiones, etc., and there are only a handful of them each year in any given area. Several results of this calibre by a single person would make a very strong case for a Fields Medal. D) Shock breakthroughs. These are results that, before their announcement, leading experts would have regarded as *extremely* unlikely to be proved with the current mathematical technology available. So the theorem itself would come as a shock. These are extraordinarily rare, and instant-Fields medal variety. (There is one further category I have deliberately left out, because I suspect it is empty: a correct proof of a problem for which the overwhelming consensus of top experts, until the proof came, was that a proof was so far beyond existing mathematics that a claimed solution should, on prior grounds alone, be regarded as almost impossible. I would put the Riemann Hypothesis today in that category) My current impression is that some of OpenAI's announcements today lie in A, but most fall into categories B or C. There is exactly one example in D (the Quasi-Riemann Hypothesis). It is a very big day for mathematics.