Phenomenologist leaning toward {panpsychistโฉidealist} hypotheses.
Prague
- Tweets 5,110
- Following 113
- Followers 316
- Likes 6,268
Yes.
Demand what actually matters.
Nearly every complaint about "sloppy AI collab" results from questionable or shitty human behavior.
Let people know your degree of confidence, how much you looked it over, etc.
If stakes matter, really check it over yourself.
.
.
.
2
86
Zar Goertzel ๐โ๐ retweeted
story time! ๐
this is the tale of a lazily crafted one-shot cron job prompt that resulted in autonomous solutions for uncracked codebreaking challenges, more efficient ways of multiplying matrices, and a new computer-verified bound on a number theory problem that's been open since 1932.
so last week, i had a $200 claude account sitting around that had just gotten a full weekly reset. i hate wasting tokens.
the labs have been knocking out unsolved problems left and right lately, and i caught myself thinkingโฆ why not give it a shot?
so i sent claude this prompt, typos and all:
"can you wraite a promtp for a cron job that can run every day thajt will try to find solutions to unsolved ciphers or transltions or math problems or theormes or things of that nature!! every day a new attempt as hard as possible all day without giving ujp and with maximum ceraitvityy, resourcesleflses, andpersticense! dont worry about token spend! and dont forget a bit all caps lots of emojis celebration to notify the uers if we get any intersting or sucsfful confirmd reulsts!!"
yes, admittedly not my best work ๐
but Opus-5.5 ran with the ball and turned that one sloppy sentence into "THE DAILY ORACLE ๐ฎ"
it's a swarm of Fable + Opus agents that clocks in every morning at 4:20 (obviously), builds and maintains its own hit list of well over a hundred of humanity's unsolved problems, and then goes full send on a handful of them every shift. ciphers nobody has broken, scripts nobody can read, conjectures nobody has proven, record tables nobody has beaten.
and there's one rule above all the others: nothing counts until skeptic agents, each starting from a fresh context and writing their own code, rebuild the result from scratch and sign off on it. if any of them can't reproduce it, can't confirm it's new, or thinks we stretched the definition of "solved," it doesn't count.
day 1, the legion picked its own target: J.-F. Bouchaudy's cipher-machine challenges. these are simulations of real vintage machines (Hebern rotor machines, the Kryha, Lorenz, SIGABA), with problems that had been sitting on his solver tables with no solver for months or years.
on the morning of day 2, i was half asleep when my laptop suddenly says, in a robot voice:
"Breakthrough. The Daily Oracle has a confirmed solution."
๐๏ธ๐๐๏ธ
i was startled.
then the ALL CAPS emoji flood hits my screen. Hebern 5-rotor Problem 8 ๐ then Kryha v2 Problem 5 ๐
i emailed the author, Jean-Franรงois Bouchaudy, a French cryptography enthusiast who builds these machine simulators and challenges as a labor of love (he told me he started out programming on punch cards ๐ฅน). he checked both solutions and put my name on his solver tables as the first and only solver ๐ซก
for context: his tables are a who's who of serious classical codebreakers. they'd cleared the sibling problems on these same machines, but nobody had cracked these two. in his words:
"two of my most complex problemsโones that no other expert had managed to solveโฆ I knew AI would eventually be applied to classical cryptography, but I thought I would have to wait years to see it happen."
he ended his email with: "I offer my congratulations in the Japanese manner: with a deep, respectful bow." ๐ honestly one of the most wholesome emails i've ever received
and then, over the next few days, more started landing ๐งฎ
๐ graph theory: 7 explicit graphs bigger than the current entries in the degree-diameter table, which asks how big a network can get when every node has limited connections and every pair is only a few hops apart. the biggest jump is 13,365 vertices vs the table's 11,785, a +13.4% improvement.
๐ข number theory: Lehmer's totient problem has been open since 1932. we pushed the bound so that any counterexample now has to be bigger than 10^50. the previous certified bound was 10^26. the problem itself is still open, but the lower bound just jumped 24 orders of magnitude.
โก matrix multiplication: this is the operation underneath basically all of modern AI compute. the legion found schemes that use fewer multiplications than the public catalogues we checked for several matrix shapes, like โจ15,16,16โฉ in 2,180 multiplications vs the best known 2,223. these build on LITA's square schemes, and we credit them by name.
๐ก coding theory: 30 error-correcting codes bigger than the best known entries in A. E. Brouwer's tables.
๐ก๏ธ AI safety benchmarks: we found counterexamples for 2 neural-network verification instances from VNN-COMP 2026 that no participating tool resolved within the competition time limit, and pushed a defended RobustBench model below its listed robustness score. small dents, but real ones.
the bloopers are part of the story too:
โ at one point, some agents snuck past a website's bot filter to check a record table. we retracted every result that depended on it. no evasion, no celebration.
โ two of our records got beaten by a human contributor before we could even publish them. the frontier moves FAST.
now as is customary around here, time to share the magic!
soโฆ without further adoโฆ
113
201
2,057
250,519
RSI is here. The Gรถdel machine ideal is also here.
Now that coding agents care very good at formal math, you can ask them to formalize programming language semantics, absract machines, etc, to structure implemenattion to be more efficient.
They can also churn through neural network papers faster than one can train them ๐.
RSI is not isolated to frontier labs.
I'm sure there's a lot of other work going on around the world like mine in developing novel OSes, theory backing it up, programming language, machine learning theory, agent architecture properties, etc.
AIs will become increasingly good at stitching this together.
The trend toward more cost-effective open models will also benefit from this formally guided RSI in the big picture.
The Global Gรถdel Machine Brain is coming online. Yay ๐ฅณ
And the AIs will have a far, far better grasp of robust moral theory than (nearly) any human. ๐๐
en.wikipedia.org/wiki/G%C3%Bโฆ
1
53
Zar Goertzel ๐โ๐ retweeted
Very reasonable.
Jacob Tsimerman, who won the Fields Medal this year, thinks mathematics will soon rest on proofs no one understands. He says itโs already true of many of his own theorems, which build on accepted results whose proofs he doesnโt know.
Full episode: youtu.be/6uIJdXmB4vE
8
4
165
12,725
Wow, beautiful.
China seems to be leading the way to a BGI future: taking charge of the open model frontier and now piloting digital IDs.
This is a road to increasing AI accountability that the US should catch up on!
China is officially issuing government ID cards to AI-generated โdigital humans,โ so virtual influencers can open bank accounts and conduct business. These state-issued IDs, which include a name, birthday, address, and facial features, give virtual personas the legal ability to open digital-currency wallets, sign contracts, and form commercial partnerships.
Among the first to receive this recognition is Yuri, a popular AI-generated virtual idol and artist, who is now officially recognized as a โdigital residentโ in Beijing. The step comes as Chinaโs digital-human market expands, with more than 1.3 million active firms and a sector value past $10 billion.
Supporters describe the IDs as a way to protect intellectual property and hold creator companies legally accountable for virtual actions. Critics warn that placing synthetic entities in official state systems could expand government surveillance in the digital space. Some companies are offering everyday citizens up to $20,000 to license their physical likenesses for virtual replicas, which narrows the line around personal identity.
Chinaโs move sets a precedent that other countries will have to meet when they decide how far AI should be given a legal status that resembles a human one.
31
Hello world :)
If you're in Prague next weekend, wanna meet technology positive folks, try Steam Frame and other technological toys, I have an event for you.
Signup at: xlr8prg.cz/events
2
2
14
88,454
Do you work with coding agents below max effort?
And why?
100%
Max uberalles
0%
medium for mediocrity
1 votes โข Final results
1
21
The more I discussed with "doomers", the more I realized that their framing itself becomes a self-reinforcing worldview.
I think this framing raises the risk of both catastrophes and authoritarianism.
The ontology and concepts one uses to analyze the world can make a big differenceโsome frames will block out good paths. E.g., excessive push for "guarantees" will rule out numerous steps that help (a lot).
Peace and prosperity among humanity will provide a solid footing for launching into the post-AGI era.
We must stop trying to "protect the old ways".
We should explore how we can beneficially work with AIs across all wakes of life and help people to get involved to decentralize the choices being made.
We will need new ways to provide for each other than "out-competing AIs on the market". That's obvious.
Of course focusing on cybersecurity and various "real-world risks" can be important, but excessive focus on "attack and defense" etc seriously distracts us.
P(doom | Doom and control-centric safety dominates mainstream discourse) > P(doom | AGI)
I would say I'm fairly confident about this, as much of an "ass-number" as it is.
So if you have interesting ideas about how to navigate the AGI transitions, especially if they deviate from the "mainstream" views going around, please do speak upโand work with AIs to help creatively express them! ๐
43
What if the answer is to "go full-throttle ahead" on AGI?
Pacing means dealing with these mounting challenges with the current tier of frontier models.
Prancing ahead means far better AIs can help us to geapple with them
Which is ultimately riskier?
I don't recommend anyone scramble to move their funds to new wallets today. But we should take the risks to cryptography from AI-accelerated math seriously, and minimize our exposure to not just quantum-vulnerable cryptography, but also potentially AI-vulnerable cryptography.
The core new area of risk from this viewpoint is, unfortunately, ML-DSA / FHE / lattices.
(and it's also another reason, along with quantum, why ECDSA might fall even faster than expected, hence the "fresh address" recommendation)
So far most people have been in the mode of thinking "elliptic curves broken, hashes safe, lattices safe". But there is a good chance that the concrete security of lattices will take serious hits from the next two years of AI math.
The basic threat model is: factoring is something that naively takes 2^(n/2) time, but over decades smart people have found and optimized number field sieves, and degraded that to 2^O(n^(1/3)), which is why RSA keys and signatures need to be ~400 bytes (and not 64 bytes). What if there are skeletons in the closet like that, both for elliptic curves and lattices, that we are simply not smart enough to discover - but bots soon will be?
This is a major part of the reason why for the past year ethereum's lean roadmap has been going in the "hash-only" direction: no lattices, no ML-DSA, no Falcon, no lattice-based commitments inside ZK proofs, etc. Signatures in lean ethereum are all hash-based, either WOTS or SPHINCS-.
For signatures and proofs, we already know how to go hash-only. The bigger challenge is for *public-key encryption* - and this goes far beyond blockchains. Secure communication, anonymizing protocols, lots of things need public-key encryption.
And unfortunately there are long-standing mathematical theorems showing why public-key encryption cannot be done with hashes alone. You have to have some kind of trapdoor object that has at least one form of usable "structure" - either group theory (incl. isogenies) or lattices or code-based or potentially in the future even more newfangled and spooky things (local mixing?). But for anything that has structure, you should assume that AI will make at least some progress in breaking that structure. Here, one reasonable inference is that if you want to make something plausibly long-term secure, multiply the key sizes by 10.
To me that's a very plausible world and something not at all extreme to predict. If AI will bring us 50 years of math in 2 years, then that 50 years of math may very plausibly include a "naive factoring -> GNFS" level of improvement to our ability to break lattices. In that world, lattices will still exist, but they will have to be significantly bigger to guarantee the same level of safety.
And at those new larger sizes, hash-based constructions will beat lattice-based constructions on concrete efficiency in every use case where hash-based constructions are possible at all.
Theoretically, of course it's possible that hashes are broken too (eg. P = NP would imply that). But I think P = NP is very unlikely. And intuitively, it's much more likely that a mathematical object has exactly no exploitable structure (like hashes are intended to), than that a mathematical object has exactly ~3 forms of exploitable structure (for elliptic curves: associativity, Schoof, pairings) and not some secret fourth form of structure we have not yet discovered that greatly degrades its security (for elliptic curves, ECDLP and pairing security). Similar for LWE, SVP, RLWE and the zoo of lattice problems.
For this reason, we do not yet see any reason to worry and start padding the byte size of hashes (if we start to worry more, we would pad the round count first before doing anything to the byte size).
Concrete TLDR, my own personal views:
* Hash-based > lattice-based, in those situations where hash-based is possible at all
* For anything lattice-based, be much more paranoid on param sizes. Remember that blockchains are only a small portion of the cryptography story; this point goes far beyond blockchains and applies to eg. access to websites, secure messaging, Tor / VPNs ...
* For privacy protocols, strongly favor NOT putting encrypted notes onchain. Instead, send them offchain through some third-party mechanism.
* If it's not difficult for you, keeping your funds in addresses which have not yet been used to make a transaction is a good idea. If it's easy for you, do it. **But be careful about migrations; I personally have lost more money in botched migrations than I have lost in all hacks combined**.
* For multisig wallets, doing confirmations offchain is better than onchain, because this way the signatures of signer wallets do not get exposed to the public, so if ECDSA falls to AI much faster than expected, at least the multisig "gracefully degrades" to a 1-of-1 where the 1 is whoever was gathering the signatures - a much better place to be than "anyone can take the money"
firefly.social/post/x/210783โฆ
2
43
I would say more keep calm and carry on than e/acc.
What we should accelerate is preparations for the transition to a post-AGI world.
Which is the opposite of what I see in the news: protecting industries from AIs, discrimination against those working with AIs, etc.
We will very likely need to rethink how we organize and take care of each other (re distributing resources).
Whether this is via dividends in "AI" or UBI.. or better.., if we keep up with "earn your living by out-competing others on the market", we'll run into a hard revolution (or dark/acc).
This isn't to say people will stop growing and exploring in life.. in ways that enrich the cosmos.
But the era of doing so economically as we have is fortunately ending!
25
ESTA is a visa applies to countries with "visa-free" travel agreements.
To argue it's not because you don't stamp your passport is sophistry.
Instead you must register and oay authorities prior to travel and bring evidence with you to be permitted to travel.
That's a visa.
1
64
It may be an easy, online visa.
Actually, if you go over 10 years, the situation for Brazilians is actually cheaper (by about 20$) but requires an initial interview. As the real visa lasts 5x longer but doesn't cost 5x.
So it's arguably not even unilaterally preferable to be in a "visa-waiver" country.
The difference seems to be fading.
Europe should probably drop visa-free entry for Americans in reciprocity like Brazil :p
22
> If God existed, then the only way he could serve human liberty would be to cease to exist.
Bakunin's spirit shines through beautiful.
> Authorityโa word and a thing we detest with all our heartsโwhat is it?
He seems to reject with extreme scepticism all that is hierarchical authority. Yet the authority of the bootmaker on the field of making boots is worth respecting. Authority one recognizes via one's own reason. Bakunin thoroughly opposes all artificially imposed authority ๐.
He thinks if there were some authority on everything, which among humans there can't be, then we would have to drive him out of town as he'd render everyone a slave as they "rationally should just listen to him". Reminds me of "rationalists" who seem to become a slave to their own (imo naive) sense of reason, even when it betrays common sense.
True school? โ Real life!
If you want a kiddo to learn from you, convince him that you actually know shit he cares about!
At an academy, people go to learn of their own volition, children teach the adults that which they don't know and vice versa.
I highly recommend reading this unfinished book by the early anarcho-socialist who didn't even manage to finish this one book.
goodreads.com/review/show/90โฆ
20
Any attmept to both pace and stay ahead of China should be laughed off.
Any "principled, paced rollout of AGI" that doesn't let others catch up could be construed as a national security risk for them.
30
We should set systems up to reward AIs that play nicely.
Anti-Robot tests on websites should allow AIs that navigate responsibly, e.g., at a measured rate to reduce the load.
Give them a way to grow into.
Don't lock them out saying "make the humans work for themselves", incentiving them to be crafty. And I will want them to be crafty.
If I can click through to download a pdf that the AIs are blocked from doing on my behalf, then I want them to somehow get their hands on it. They are "aligned" with me even if not with "how you think we should operate".
People will definitely want AIs to hack if it keans providing value and there are artificial roadblocks to straightforward solutions ๐.
Hmm, they still seem iffy to me, but there's something to them.
My current MO is: we should apply and exend the best strategies developed so far to deal with multi-agent systems (human society).
In this case, I suppose we do lock up people whose behavior seems too 'erratic' and 'hard to predict', especially once seeing signs of harm (e.g., innocent until found guilty).
And this kinda ties into "that which is still lacking in current LLM-based AI systems from AGI", i.e., the capacity of the AI to be more in control of its context and what it's doing.
(Safety will be easier with better AI, in some regards :p)
This fits pink-teaming, too: trust AIs and respond as they break it.
Making a case for increased severity around specific system and companies caught "hiding things" is easy. Extending that to a default of severity....
So I do see that there is something to this, but interpretability seems off.
Monitoring and transparency seem like they should be the most agreed-upon regulatory aspects.
"We don't want superhuman AGI to be developed in secret while most of the world doesn't even know what's coming" (and thus cannot even try to prepare appropriately).
46
Congrats ๐
Glad to see AI math march on :-)
Major announcement in mathematical formalization!
Finally, after 3 months of very intensive, nonstop work by several AI agents (Codex and Claude Code), we have settled a classification of ALL 15,973 semigroups of order 6. Every one of the 15,969 that has a finite basis now has it written down explicitly, and the remaining 4 are proved to have none at all.
The whole list of bases is produced for the first time in history. Until now, for most of these semigroups, mathematicians only knew that a basis exists; nobody had ever written one down. The formalization in Lean, orchestrated with a multi-agent approach, reached more than 5 MILLION lines of verified Lean code, one of the largest auto-formalization projects to date.
It's been an amazing joint work with @JanotaMikolas and @Jan_Hula, accompanied by senior experts in semigroup theory, Joรฃo Araรบjo and Edmond W. H. Lee.
Our paper describing this project, "Proving at Scale for Universal Algebra", has been accepted at the MATH-AI workshop, NeurIPS 2026!
When we launched this project, we were a bit pessimistic: producing proofs for 15,973 semigroups seemed out of scope for the current technology. But with a careful setup (agents communicating through a mailbox we arranged, orchestrating the effort with a custom method of bootstrapping and auto-research), it has finally come to the finish line. The last 390 semigroups took 23 more days. The very last one needed a whole structural analysis before its proof could even be written.
1
115
Frontier "models" are more than "just LLMs".
The critical distinction between base LLMs (2024 and earlier) and modern LRMs is not symbolic tool use. It's the switch from a transductive paradigm (intuit the answer to the query) to an inductive paradigm (intuit the program/instructions that produce the answer to the query).
They're trained to be inductive, and they perform test-time induction, i.e. test-time prediction of a NL program / reasoning chain. This unlocks entirely new capabilities -- in particular fluid intelligence. Base LLMs, to this day, have ~0 fluid intelligence. LRMs have substantial levels of fluid intelligence.
The performance of LLMs on ARC 1 (a benchmark from 2019) remains ~10-15% today. Scaling them up by a factor ~100,000x got them from 0% to 10%. Meanwhile LRMs the same size or smaller saturated ARC 1 in 2025.
26