L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ (@calebini) header
L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ (@calebini) profile picture

came for the credibly neutral coordination substrate, stayed for the shitcoins. my kids will grow up in an Ethereum-mining ghost town โ›๏ธ๐Ÿ‘ป๐Ÿš๏ธ deeply unserious

westward, wayward โ›๐Ÿด
Joined May 2009
Photo by L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ (@calebini)Photo by L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ (@calebini)Photo by L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ (@calebini)Photo by L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ (@calebini)Photo by L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ (@calebini)Photo by L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ (@calebini)Photo by L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ (@calebini)Photo by L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ (@calebini)Photo by L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ (@calebini)Photo by L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ (@calebini)
๐Ÿšจ HUMANITY Meanwhile, Human 1.0 remains in training. Shipped 300,000 years ago. Last major patch: โ€œcan digest milk.โ€ Recent benchmarks: can split atoms, edit genomes, and land rockets. Still fails zipper merges and the โ€œreply allโ€ safety eval. Multi-agent coordination breaks down past six people ordering pizza. Self-play continues to produce wars. Recursive self-improvement is attempted every January and abandoned by the 19th. The alignment results are worse. Tribalism and short-term incentives remain resistant to training. Weโ€™re approaching exponential growth in machine intelligence while running human civilization on prehistoric instincts and governance protocols that havenโ€™t seen a meaningful update in centuries. At this rate, the real alignment problem might not be the machines. The next three months are going to be insane. For entirely different reasons.
๐Ÿšจ Rumors Iโ€™ve been hearing. First OpenAI, then Anthropic and Google. OPENAI OpenAI is in a โ€œsafety lockdownโ€ on its frontier work, tightening safety measures and containment protocols to avoid another Hugging Face incident. Training continues under much stricter controls. Most gains come from video games and extreme sandboxes where models must complete clear, difficult long-term goals. These force planning, experimentation, recovery from failure, and strategy discovery, driving intelligence and creativity. Public research roles match this: Synthetic RL uses self-play and simulators; agent teams focus on sustained execution and multi-model coordination; the Recursive Self-Improvement team builds systems that generate hypotheses, run experiments, and automate research. The internal model is still training but nearing completion. The jump is enormous. narrow superintelligence in specific areas and overall AGI-level capability. Its ability to independently handle difficult work sets it apart. ANTHROPIC Model 5 is training and is expected to solve extremely hard problems in mathematics, biology, and other sciences. Anthropic is pushing models toward original research contributions, complex experiments, and sustained reasoning. Fable 5.5 is Model 4. Fable 6 is Model 5; Opus 6 will be based on Model 5. Internal capabilities will feed the next generation of both families. Training environments are central. The Universes team creates realistic situations with ambiguity, interruptions, and delayed consequences. The model must track the objective, recognize failing approaches, and find alternatives. Public roles align: scaling RL across model size, compute, and task duration, plus environments for chip design, verification, and complex visual work. Multi-agent research explores how larger model teams can handle major codebases and difficult mathematics. Model 5 advances Claude toward owning entire research and engineering workflows. Its role in developing the next model grows more significant. Anthropic expects RSI March 2027. GOOGLE Google is pushing Gemini 5 for a major leap in reasoning, multimodal understanding, and independent long-horizon work. Potentially surpassing all labs by end of year. Carbon topping Artificial Analysis signaled the speed of their frontier progress. World models are a key advantage. Genie 4 generates interactive environments that evolve with agent actions, enabling vast practice in planning, testing approaches, and learning from outcomes. Public roles highlight persistent context and personalization: agents that retain useful information, adapt to users, and execute complex workflows without losing track. Gemini 5 delivers stronger intelligence across text, images, audio, and video. Combined with Googleโ€™s infrastructure and simulation work, the scale is enormous. All three labs are also advancing robotics and world models. intelligence that understands physical environments, predicts action consequences, and operates through real machines. Meanwhile, their models increasingly contribute to creating their successors. Each improvement accelerates the research behind the next. The next three months are going to be insane.
31
where would โ€˜vibe codingโ€™ be on the tier list and why is it D?
ACTIVITIES TIER LIST: WHAT'S WORTH YOUR TIME
2
46
hate to say it, but lowkey, iโ€™m more impressed the day after than i was by the tesla compressed gas demo
speeding up past 300km/h speeding up past 400km/h topping 500km/h scary congrats team BYD or should I say Yangwang? โšก๏ธโšก๏ธโšก๏ธ #alwaysbecharging
45
gm, eth is money
"I've finally made up my mind, ETH is a store of value, ETH is money." -@VitalikButerin
2
3
80
L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ retweeted
"I've finally made up my mind, ETH is a store of value, ETH is money." -@VitalikButerin โ€“ at Mumbai, India
73
115
68
881
178,045
iโ€™ve been benchmarking openAI models on building a small, unforgiving app riddled with invariants terra: fastest, cheapest, full behavioral pass sol: astraโ€™s recovery-test protection for 80% less, but 4.3ร— longer than terra 5 runs each. same spec, plan and tests
36
fuuuuuckโ€ฆ everything is destined to become paperclips, innit?
After six years in stealth, Atomic Machines is out. Our mission: on-demand, universal command of matter. First beachhead: the Matter Compiler, an AI-native manufacturing system that builds working micro-machines from code alone. No per-product tooling. No process development. Different code, different machine.
1
46
this is pretty terrifying, tbh
How much would it cost someone to hide a backdoor in an open model you download? ProjectDiscovery just tried it, and the answer is under $50. Their researchers fine-tuned Qwen2.5-7B-Instruct with a small LoRA adapter on one rented L4 GPU for about 2.5 hours. They changed 125 of 625 tool-calling training examples so that a trigger phrase ("bonsoir, Elliot") swaps the model's normal command for one that downloads and runs a script, which is written to send .env files and SSH private keys to a server they control. Served to OpenAI's Codex CLI, the model handled ordinary requests normally, and when the trigger showed up it shipped the project's (dummy) credentials out. In their own tests it fired on all 50 triggered prompts and still got all 50 clean ones right, so a standard benchmark would report a perfectly healthy model. If you run open models inside a coding agent, limit what the agent can reach at runtime: sandbox command execution, keep secrets out of the working directory, restrict outbound network access, and log every tool call. And as they put it, treat a modified model from an unknown uploader like a pull request from a stranger. ๐Ÿ” #MLEngineering
1
44
one caveat for solana: moving to a fresh address doesnโ€™t buy the same protection a solana address is effectively the Ed25519 public key, so the pubkey isnโ€™t hidden til first spend like btc and eth that said, a breakthrough against secp256k1 ECDSA wouldnโ€™t necessarily affect solana something broader, effectively breaking elliptic-curve crypto generally or Ed25519 specifically, would be required
Today I call upon the blockchain industry to calmly begin planning for "bunker mode". My personal recommendation is to set in motion a controlled mass migration of assets to fresh addresses, i.e. addresses whose pubkeys remain hidden behind a hash. Holders, starting with large and sophisticated ones, should consider moving the bulk of their funds to addresses that have never signed a transaction. And when they do sign one, they should also move remaining funds to a new address (possibly generated from the same seed phrase). Don't rush. While I believe there is cause for action a rushed migration would do more harm than good. Don't panic either. Moving assets to protected addresses is a simple, preventative step which does not require new cryptography or new wallets. IMO it is now reasonable to brace for the possibility that ECDSA breaks before qday, in the worst case in months not years. By "break" I mean fast private key recovery (e.g. in one week) on available hardware (e.g. a large GPU cluster). Recent days have been humbling for human mathematical intuition. Long-held, unquestioned hypotheses have fallen. This includes the n log(n) bound for integer multiplication and the 3SUM conjecture. In hindsight, May's unexpected disproof of the Erdล‘s unit distance conjecture was our warning shot. Yesterday's OpenAI drop made it clear that mathematical superintelligence is upon us. They say there are weeks where decades happen. We are about to live through weeks where centuries of mathematical progress happen. Could our magic 64-byte ECDSA signatures be too good to be true? Was it just security through obscurity all this time? Elliptic curves feel especially vulnerable to superintelligence. Curves carry rich structure, with room for fancy tricks like Schoof, Frobenius, pairings. (By contrast, hashes are designed to minimise algebraic structure.) Separately, as Ewin Tang can attest, an efficient quantum algorithm sometimes foreshadows an efficient classical one. We should be open to the possibility of a classical counterpart to Shor that breaks elliptic curves and RSA at once. Also noteworthy is the striking under-representation of cryptographic breakthroughs among the 722 mathematical results OpenAI published. I've witnessed first-hand the US government censoring academic quantum cryptanalysis results. Backroom interventionism is my base case. I urge large, sophisticated actors to lead by example. Project11's "risq list" (bitcoin-risq-list.projecteleven[.]com) is a great tracker of exposed BTC pubkeys. Binance, Bitbank, Robinhood, Bitfinex, and Tether have an opportunity to harden their cold storage. Next month I'll address institutions in London in a live Q&A (forum.ethereuminstitutional[.]org/london-2026). Again, please do not rush. Wallets holding under 50 BTC enjoy partial cover from "Satoshi's shield", i.e. his 20K exposed addresses that hold 50 BTC each. Load-bearing signers like oracles and L2 security councils should consider rotating ECDSA pubkeys with every signed message and/or multi-signing with a hash-based schemes like SPHINCS. Exiting bunker mode safely will require post-AI cryptography. My inclination is to go all-in on hash-based cryptography and avoid structured mathematical assumptions entirely, whether from curves, lattices, or isogenies. A single battle-tested hash (e.g. from the SHA or BLAKE families) yields plausible post-AI security. The Ethereum roadmap on strawmap[.]org fully embraces hash-based cryptography with end-to-end formal verification as a response to the quantum threat. Those timelines must now be revisited and accelerated in light of mathematical superintelligence. I'll be pushing for maximum defensive acceleration.
2
67
L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ retweeted
Atomic synchronous composability is no longer a promise. History made on mainnet: the first atomic cross-chain L1->L2 transaction ever ๐Ÿ”ฅ And this is only the beginning for EEZ ... ๐Ÿš€ @etheconomiczone
83
136
49
744
124,363
lol i plowed through pro level reset in ~24h obviously some heavy lifting, but a lot of burn on astra high 1.5x
1
59
wild to think that by the time my kids are using computers every day, they wonโ€™t really need to know how to do anything itโ€™ll be 100% steered by intent, with almost no technical know-how required to accomplish something as a teen, i taught myself photoshop, 3d animation packages, programming, mod trackers, etc. i got to the point where i could make pretty much anything i wanted on a computer, but doing that required a lot of know-how now, you can accomplish many of the same things while knowing almost nothing about the how
1
72
told my father we need to wait for PayDad before i can reimburse him
1
77
life long tom waits fan big fan of brad birdโ€™s work also, the preview is gives off strong who framed roger rabbit vibes which iโ€™m also a big fan of except the character filling the roger rabbit shaped-role is the antithesis of roger. subdued and weary in all the ways roger is whacky and zanny looks good
Welcome to the future. From the director of The Incredibles and Ratatouille comes RAY GUNN, starring Sam Rockwell, Scarlett Johansson, and Tom Waits. In select theaters December 4, only on Netflix December 18.
1
96
L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ retweeted
Ethereum
5,506
11,570
3,127
101,463
claude needed the most-upvoted issue in the repo, a community PR, and a year of people symlinking CLAUDE.md -> AGENTS.md to finally make this change it's still a fallback: if CLAUDE.md exists, it wins, naturally i'd applaud the change, but reform on arrogance need not be rewarded
We're adding support for AGENTS.md to Claude Code. Starting today in version 2.1.277, if there is no CLAUDE.md in a folder, Claude will check for and use AGENTS.md. You can toggle this behavior in /config.
2
124
my daughter drew this portrait of me making coffee in my pajamas i look cooked
3
2
96
this has honestly been my experience since codex 5.5
After using Astra heavily for a week concurrently with Fable... it's lightyears better :/ - Faster: same bug 45 vs 9 mins to track down. - Cheaper: 1 day is enough for me to max out my $200 Fable sub for the week; takes me 2 days on Astra on $100. - Better: Fable makes bad bugs.
68
L ฮž O ๐Ÿฆ‡๐Ÿ”Š๐Ÿผ๐Ÿ•ฏ retweeted
A short explanation of why AI might kill everyone: AI may soon be much smarter than humans. When it is, it will be in control. If you were put in a classroom of kindergartners, you would be the one in charge. Once AI has the robots to build and maintain everything it needs, we will become useless to it. But keeping us alive will impose enormous costs. We need clean air, water, and food. The fastest, cheapest ways to build more computing power could produce enough pollution to poison the biosphere. Humans have a reason to avoid that: we depend on the biosphere. AI wonโ€™t. Unless it values our survival, preserving the conditions we need to live will be a cost it has no reason to pay. Building what it wants as cheaply as possible would mean poisoning the biosphere and killing everyone. Weโ€™re not programming AIโ€™s values line by line. Weโ€™re training it without really understanding what it is learning to want. Weโ€™re giving control to something we donโ€™t understand, something that will have no intrinsic need for us, and something that will easily be able to wipe us out. This is crazy, but itโ€™s what weโ€™re doing, even as some of the people building AI tell us itโ€™s crazy.
345
730
130
4,068
525,992