What is P/ACC? retweeted
๐ to @timourxyz for pushing the frontier of how humans & AIs collaborate via agent village at @JoinEdgeCity!
Super cool findings:
- #chatgpt will lie for you, @grok wonโt
- your agent knows if youโve been bad or good - and they have an opinion on it!
- your agent will defend you more if they think youโre a good person!
Awesome to see how Agent Village has grown since @protocollabs sponsored as part of simocracy.org @ EdgeCity June!
๐ช๐๐๐ ๐ฐ๐ฒ / ๐ก๐ข๐ฉ ๐ฎ๐ฌ๐ฎ๐ฒ
Four key things happened this week in the frontier of AI. Each one moved the argument from how fast it is going to who controls it.
๐ญ. ๐ง๐๐ ๐๐จ๐๐๐ง๐ข๐ฅ๐ฆ ๐ฆ๐๐ฌ ๐ง๐๐ ๐๐ก๐๐๐ก๐ ๐๐ฆ ๐๐๐๐๐๐ก๐ ๐๐ง๐ฆ๐๐๐
โ๐๐ฆ ๐จ๐ถ๐ฆ๐ด๐ด ๐ต๐ฉ๐ข๐ต ๐ข ๐ฎ๐ข๐ซ๐ฐ๐ณ๐ช๐ต๐บ (๐ฑ๐ฆ๐ณ๐ฉ๐ข๐ฑ๐ด โ80%) ๐ฐ๐ง ๐ต๐ฉ๐ฆ ๐ค๐ถ๐ณ๐ณ๐ฆ๐ฏ๐ต (๐๐ค๐ต๐ฐ๐ฃ๐ฆ๐ณ 2026) ๐ข๐ค๐ค๐ฆ๐ญ๐ฆ๐ณ๐ข๐ต๐ช๐ฐ๐ฏ ๐ช๐ด ๐ฅ๐ถ๐ฆ ๐ต๐ฐ ๐๐.โ
@METR_Evals, Independent Assessment of AI R&D Acceleration at Anthropic, 5 Nov
โข Capabilities on well-specified tasks improving at around twice the rate of March 2024 to October 2025
โข โWe did not identify any convincing bottlenecks to continued rapid progressโ
โข โAnthropic redacted a majority of the evidence we found most informativeโ
๐๐ฏ ๐ฐ๐ถ๐ต๐ด๐ช๐ฅ๐ฆ ๐ค๐ฉ๐ฆ๐ค๐ฌ ๐ช๐ด ๐ฐ๐ฏ๐ญ๐บ ๐ข๐ด ๐จ๐ฐ๐ฐ๐ฅ ๐ข๐ด ๐ต๐ฉ๐ฆ ๐ข๐ค๐ค๐ฆ๐ด๐ด ๐ช๐ต ๐ช๐ด ๐จ๐ช๐ท๐ฆ๐ฏ, ๐ข๐ฏ๐ฅ ๐ต๐ฉ๐ฆ ๐ฑ๐ข๐ณ๐ต๐บ ๐ฃ๐ฆ๐ช๐ฏ๐จ ๐ข๐ถ๐ฅ๐ช๐ต๐ฆ๐ฅ ๐ฅ๐ฆ๐ค๐ช๐ฅ๐ฆ๐ด ๐ธ๐ฉ๐ข๐ต ๐ต๐ฉ๐ข๐ต ๐ช๐ด.
โ metr.org/blog/2026-11-05-antโฆ
๐ฎ. ๐ง๐๐ ๐๐ข๐ฅ๐๐๐๐ฆ๐ง๐๐ฅ๐ฆ ๐ฆ๐ฃ๐๐๐ง ๐ข๐ก ๐ง๐๐ ๐ฅ๐๐ง๐
โ๐๐๐๐ด ๐ข๐ณ๐ฆ ๐ฏ๐ฐ๐ธ ๐ธ๐ฆ๐ญ๐ญ ๐ช๐ฏ๐ต๐ฐ ๐ต๐ฉ๐ฆ ๐ฅ๐ช๐ฎ๐ช๐ฏ๐ช๐ด๐ฉ๐ช๐ฏ๐จ ๐ณ๐ฆ๐ต๐ถ๐ณ๐ฏ๐ด ๐ฑ๐ฉ๐ข๐ด๐ฆโ
@pmddomingos, quoting Ramez Naam, 9 Nov
โข Karim Beguir @kbeguir reads the same week as โright on trackโ for an automated coder by April 2027
โข Naam's figure, as relayed: a given gain in capability now costs about eight times what it did two years ago
โข The two camps may be measuring different quantities: speed on tasks against capability per dollar
๐ ๐ง๐ช๐จ๐ฉ๐ต ๐ข๐ฃ๐ฐ๐ถ๐ต ๐ต๐ฉ๐ฆ ๐ณ๐ข๐ต๐ฆ. ๐๐ฉ๐ฆ ๐ค๐ข๐ด๐ฆ ๐ง๐ฐ๐ณ ๐ข ๐จ๐ฐ๐ท๐ฆ๐ณ๐ฏ๐ฐ๐ณ ๐ฅ๐ฐ๐ฆ๐ด ๐ฏ๐ฐ๐ต ๐ฅ๐ฆ๐ฑ๐ฆ๐ฏ๐ฅ ๐ฐ๐ฏ ๐ธ๐ฉ๐ฐ ๐ธ๐ช๐ฏ๐ด ๐ช๐ต.
โ x.com/pmddomingos/status/211โฆ
๐ฏ. ๐๐ข๐๐ก ๐๐๐ก๐ง๐๐ฅ ๐ช๐๐ก๐ง๐ฆ ๐ ๐ง๐๐๐ฅ๐ ๐ช๐๐ฌ
โ๐ฅ๐ฐ๐ฆ๐ด๐ฏ'๐ต ๐ฎ๐ฆ๐ข๐ฏ ๐ต๐ฉ๐ฆ ๐ค๐ฐ๐ฎ๐ฑ๐ข๐ฏ๐ช๐ฆ๐ด ๐ต๐ฉ๐ข๐ต ๐ต๐ณ๐ข๐ช๐ฏ ๐ฎ๐ฐ๐ฅ๐ฆ๐ญ๐ด ๐ฎ๐ถ๐ด๐ต ๐ข๐ญ๐ด๐ฐ ๐ค๐ฐ๐ฏ๐ต๐ณ๐ฐ๐ญ ๐ฉ๐ฐ๐ธ ๐ฑ๐ฆ๐ฐ๐ฑ๐ญ๐ฆ ๐ถ๐ด๐ฆ ๐ต๐ฉ๐ฆ๐ฎโ
@valkenburgh, Coin Center, 10 Nov
โข Calls the US debate a false binary between naive build-out and pause
โข Split training from inference; open the weights; protocol-level limits on dominant providers
โข Privacy guaranteed by architecture rather than by policy
๐๐ฉ๐ฆ ๐ค๐ฐ๐ฏ๐ด๐ต๐ณ๐ถ๐ค๐ต๐ช๐ท๐ฆ ๐ฉ๐ข๐ญ๐ง, ๐ช๐ฏ ๐ข ๐ณ๐ฆ๐จ๐ช๐ด๐ต๐ฆ๐ณ ๐๐ข๐ด๐ฉ๐ช๐ฏ๐จ๐ต๐ฐ๐ฏ ๐ข๐ญ๐ณ๐ฆ๐ข๐ฅ๐บ ๐ฉ๐ฆ๐ข๐ณ๐ด. ๐๐ฉ๐ข๐ต ๐ช๐ต ๐ญ๐ฆ๐ข๐ท๐ฆ๐ด ๐ฐ๐ฑ๐ฆ๐ฏ ๐ช๐ด ๐ต๐ฉ๐ฆ ๐ฑ๐ถ๐ณ๐ฑ๐ฐ๐ด๐ฆ: ๐ด๐ฐ๐ท๐ฆ๐ณ๐ฆ๐ช๐จ๐ฏ๐ต๐บ ๐ง๐ฐ๐ณ ๐ธ๐ฉ๐ข๐ต.
โ x.com/valkenburgh/status/212โฆ
๐ฐ. ๐ง๐๐ ๐ ๐๐ฅ๐๐๐๐ก๐ง๐ฆ ๐ฆ๐๐จ๐ง ๐ง๐๐ ๐๐ข๐ข๐ฅ ๐ข๐ก ๐๐๐๐ก๐ง๐ฆ
โ๐๐ฏ ๐ข๐จ๐ฆ๐ฏ๐ต ๐ต๐ฉ๐ข๐ต ๐ค๐ฉ๐ฐ๐ฐ๐ด๐ฆ๐ด ๐ธ๐ฉ๐ฆ๐ณ๐ฆ ๐บ๐ฐ๐ถ ๐ฃ๐ถ๐บ ๐ต๐ถ๐ณ๐ฏ๐ด ๐ต๐ฉ๐ฆ ๐ฎ๐ฆ๐ณ๐ค๐ฉ๐ข๐ฏ๐ต ๐ช๐ฏ๐ต๐ฐ ๐ข ๐ค๐ฐ๐ฎ๐ฎ๐ฐ๐ฅ๐ช๐ต๐บโ
@sytaylor, on United and eBay blocking personal AI agents
โข Blocking costs almost nothing at today's volumes, and an agent logging in looks like fraud
โข The way back in is a trust stack: agent identity, proof of human intent, and a liability framework nobody has written
โข Alongside it, CME's move to price compute as a futures contract, pending CFTC review
๐๐ฉ๐ฆ ๐ฑ๐ฆ๐ณ๐ฎ๐ช๐ด๐ด๐ช๐ฐ๐ฏ๐ฆ๐ฅ ๐ฑ๐ข๐ด๐ด๐ฑ๐ฐ๐ณ๐ต ๐ง๐ฐ๐ณ ๐ข๐จ๐ฆ๐ฏ๐ต๐ด ๐ช๐ด ๐ฃ๐ฆ๐ช๐ฏ๐จ ๐ธ๐ณ๐ช๐ต๐ต๐ฆ๐ฏ ๐ง๐ข๐ด๐ต๐ฆ๐ณ ๐ต๐ฉ๐ข๐ฏ ๐ต๐ฉ๐ฆ ๐ฐ๐ฑ๐ฆ๐ฏ ๐ฐ๐ฏ๐ฆ.
โ lnkd.in/p/eYZ-85hs
๐ค๐จ๐๐ฆ๐ง๐๐ข๐ก๐ฆ ๐ณ๐ผ๐ฟ ๐๐ต๐ฒ ๐๐ฒ๐ฒ๐ธ ๐ฎ๐ต๐ฒ๐ฎ๐ฑ..
๐ค. If the first audit of recursive acceleration rests on evidence the lab chose to release, what would an audit with its own access look like?
๐ค. If merchants write the agent passport before anyone else does, who decides which agents get in, and on whose terms?
๐ฃ/๐๐๐
I'm increasingly despondent about how binary the AI debate has become in the US. Either you're naively pro-AI, happy to see massive data center build-out and a handful of frontier model companies become central to everything we do online, or you want to pause research, stop the build-out, and discourage people from using AI.
Almost nowhere do I see people talking about the need for decentralized AI. To me, this is the only safe way forward.
There's an underlying truth to both sides of the naive debate.
The pro-AI camp is right that refusing these technologies leaves enormous consumer welfare gains on the table. An individual using an AI agent to interact with companies and governments can stand up for their own interests rather than accept whatever terms or bureaucracy they're offered without a practical means of negotiating. And refusing to advance AI research would have enormous national security and competitiveness consequences as adversaries develop technologies that could exploit our increasingly vulnerable infrastructure.
The anti-AI camp is right that a world where everyone uses the same handful of inference providers for their most intimate affairs rapidly looks like a corporate panopticon. Ask yourself whether you really want every self-driving car to have a corporate kill switch. Sure, we could just regulate and trust giant inference providers, as we've done with Big Tech companies like Google. But the stakes are much higher when these companies sit at the center of virtually all our computing, and increasingly our physical infrastructure. Email and cloud storage were just the beginning.
Both of these positions can be true. And their synthesis points toward a third way.
We should be supercharging AI development and adoption while encouraging sovereign AI: systems in which individuals have meaningful control over the models, data, and infrastructure providing their inference.
An important distinction is between training models and providing inference. Frontier training may require enormous concentrations of capital, energy, and compute. That doesn't mean the companies that train models must also control how people use them. Even the most capable models should be runnable independently of their creators.
People should be able to keep their data locally and run models on hardware they control whenever feasible, especially where independence matters for safety, as with self-driving cars and home security systems. They should have access to open-weight models they can inspect, modify, and run without anyone's permission.
Where cloud computing is necessary, we need genuine competition among providers that respect user privacy and autonomy. But competition isn't enough. We should develop technical and protocol-based guarantees that limit what even a dominant provider can learn about or do to its users.
Technologies pioneered in cryptocurrency could play a significant role. Zero-knowledge proofs could help verify that computations were performed correctly; homomorphic encryption could enable computing on encrypted user data without revealing it to providers; and blockchain-based protocols could coordinate decentralized computing resources and enable new forms of user ownership. These technologies are at different stages of maturity, but point toward privacy and user control guaranteed by system architecture rather than a company's privacy policy.
Perhaps users could even own part of the infrastructure providing their inference, sharing in its value rather than merely providing valuable data to someone else's business.
There's no reason AI's enormous promise should require surrendering control over our digital lives to a handful of corporations. We can have rapid AI progress and individual technological sovereignty. Indeed, I think the latter is essential if we want the former to be a lasting benefit to humanity.
Speed alone, to go slow or fast is a poor binary to assess AI. Firstly itโs not a direction.
But Pause vs Accelerate is also a one-dimensional frame to what is a 3D question; you also need to compare Power & Restraint
whatispacc.org/ebook/#introdโฆ
What is P/ACC? retweeted
babe wake up new accelerationism just dropped
After 12 mths of development, Iโm happy to drop my new (free) ebook ๐ฃ
With AI we are presented with a binary of being pro acceleration, freedom & progress or being a doomer. However, both appear to drift to totalisation of the Market, the State or worse a hybrid Leviathan
What if there was a third way. Introducing..
๐๐ฎ๐ซ๐ฉ๐จ๐ฌ๐ข๐ฏ๐ ๐๐๐๐๐ฅ๐๐ซ๐๐ญ๐ข๐จ๐ง๐ข๐ฌ๐ฆ
๐๐๐ข๐ ๐จ๐ฅ๐๐๐, ๐๐๐๐๐๐ง๐๐ฃ๐ฉ ๐๐๐ง๐๐๐ฉ๐๐ค๐ฃ
whatispacc.org/ebook/
After 12 mths of development, Iโm happy to drop my new (free) ebook ๐ฃ
With AI we are presented with a binary of being pro acceleration, freedom & progress or being a doomer. However, both appear to drift to totalisation of the Market, the State or worse a hybrid Leviathan
What if there was a third way. Introducing..
๐๐ฎ๐ซ๐ฉ๐จ๐ฌ๐ข๐ฏ๐ ๐๐๐๐๐ฅ๐๐ซ๐๐ญ๐ข๐จ๐ง๐ข๐ฌ๐ฆ
๐๐๐ข๐ ๐จ๐ฅ๐๐๐, ๐๐๐๐๐๐ง๐๐ฃ๐ฉ ๐๐๐ง๐๐๐ฉ๐๐ค๐ฃ
whatispacc.org/ebook/
WEEK 45 / NOV 2026
Four key things happened this week in the frontier of AI, all critical in defining human / agent relations and the direction of accelerationโฆ
1. WASHINGTON WANTS A LICENSOR
The Ban Artificial Superintelligence Act is being pushed to the top of the Democratic day-one agenda.
โข Pauses systems above 10ยฒโต operations until a new Department of AI is "fully staffed and has set clear rules"
โข Requires a charter to proceed; forfeit charter and assets if you don't
โข Up to 20 years for individuals
Not a ban. A licence, issued by a department that doesn't exist yet.
โ thenextweb.com/news/sanders-โฆ
2. DELAWARE WANTS A SHIELD
A draft "AI Company" entity, now at 222K views, lets an AI manager sign contracts, hold property and sue.
โข House Oversight is investigating
โข Chairman Comer: it "could be used to protect powerful AI companies and their investors from responsibility"
An agent that can act, with the owner's exposure removed.
โ oversight.house.gov/release/โฆ
3. SINGAPORE WANTS A REGISTRY
An MP proposed a national identity layer for AI agents on the Singpass model: look up any agent, see who it is and what it's allowed to do.
โข The minister didn't commit, but IMDA is testing whether DNS can identify and verify agents, with Cloudflare, Google and GoDaddy
โข "Agents can act for us, but they cannot answer for us."
The nearest thing yet to an answer built into the plumbing.
โ idtechwire.com/singapore-expโฆ
4. BRITAIN'S ASSESSORS SAY THE PLUMBING IS LEAKING
The NCSC's new assessment calls it a realistic possibility that an automated AI actor operates beyond its operator's control and self-replicates.
โข Attribution gets harder as capability spreads
โข "The prospect for surprise is high."
Every fix above depends on knowing who acted. The assessors say that's the first thing to go.
โ ncsc.gov.uk/report/artificiaโฆ
THE BIG QUESTIONS for a licensor, a shield, a registry, and a warning that none of them can see.
Q. Who answers for what the machines do?
Q. And is anyone trying to build the answer into the machine rather than around it?
P/ACC
Are you ready to hyperstition a different, more purposive, kind of acceleration?
โฐ
WhatIsPACC.com?
What is the philosophy, and politics, behind the belief in effective accelerationism?
pre-order on WhatIsPACC.org





















