Millions of robots 🤖 live for milli-seconds⚡to make the markets turing complete , who will be the last human 👥 standing? I live till I decentralize. ⚓🏴☠️
SPV for Surplus Human Capital Mining:
Flash Organization: Flashmob w/ phase transition b/t open community & value-capturing temp constructs👷
Application Specific Task Force: ASIC of heterogeneous capital formation w/ a count-down timer🏴☠️
MicroDAO: microbrewery of DAOs🐄
Excellent news: The Peraire-Bueno brothers have been acquitted. Congratulations to them and their defense teams!
Coin Center filed an amicus brief challenging the SDNY’s “honest validator” theory of wire fraud. In Ethereum, “honest” validation has a specific technical meaning. It does not create an open-ended legal duty to meet prosecutors’ expectations of fair play.
Had the SDNY’s theory prevailed, the government could use wire fraud law to impose its own code of conduct on Ethereum validators, backed by criminal penalties. Anyone participating in an open network would have to wonder whether prosecutors might later declare their behavior “dishonest” and haul them into court.
[ ZOOMER ]
JUDGE RULES $25M MEV HEIST THAT HONEYPOTTED SANDWICH BOTS WAS NOT WIREFRAUD AS IT DID NOT HAPPEN ON A REGULATED MARKET, ACQUITS MIT BROTHERS: RTRS
Searching for open-weight models can be so much better.
The right model for you shouldn’t be buried in a directory. The models must flow.
Introducing model trees. Check the revamped models page ↓
Featuring @elder_plinius@support_huihui@QwenDevs
Upgrading a stateful app in a dstack TEE without downtime: boot a second CVM with the same app_id, hand state over with a GetKey-sealed handover, delete the old one. On Phala: in-place update = 65 s outage; double-buffered = 0 failed of 503. github.com/Dstack-TEE/dstack…
Time to revisit the Tower of Babel.
where complexity theory, limits of physics, complicates the notion of shared reality.
ebb and flow of the ephemeral social fabric of "trust",
inevitably warps any quests for "human agency" that slips our grasp like Schrodinger's cat.
Founders are starting to ship a company.md with their pitches: a markdown file written for AI consumption (alongside deck and data room).
As a VC, I consider this a very positive signal.
Instead of force-feeding my agent excel sheets, PDFs, and screenshots of charts, I can hand it one artifact that compresses everything I should know about your business.
What it tells me:
1) You respect my time. I can debate, discuss, and probe your business before we meet, so we can skip the "wait, wtf are you doing?" part and get straight to the heart of what you're building.
2) You understand how investors work now.
3) You're AI-native.
4) You're not afraid of a neutral read on your company. My AI will give me that anyway, the only question is whether you'll make me do more steps in between to get there.
One warning: I will ask my AI how biased your company.md was, and if there's anything in there that feels misleading. The more objective, the better.
This should become the gold standard for fundraising.
A binding commitment to a future action restricts your choices (intentionally).
In a strategic situation (game, market, etc), those restrictions on your choices change the game tree, which can result in new equilibria that didn’t exist before your commitment.
Excited to see where this goes — between abliteration and "there are so many random techniques we haven't tried yet to edit at the weight level yet," it feels like there's a long tail of low-level optimizations to be explored (and exploring them via contest is the fastest route)
Releasing Delta League: edit LLM weights for benchmarking and glory! Join at deltaleague(.)ai
Earlier this week, the idea of editing weights directly with an LLM got a lot of attention —
but it was shunted by limitations in the demonstrated approach.
Delta League is an open competition for everyone to submit deltas to model weights and get them scored on a hidden test set. Auto-research and hillclimb to make smaller models more performant.
The broader vision for Delta League is creating a platform for open AI research. Beyond just editing weights, this approach offers verifiable evaluation and an open collaborative space for everyone to continue improving Pareto of open source models.
In the coming days, we will extend this platform to support editing model architectures beyond just weights and custom evaluation sets to understand how models perform in specific domains and for businesses.
Yes, that's correct. New fabs need years and billions to build and ramp, so capacity is effectively fixed in the near term. AI demand grew faster than the industry planned for, locking most HBM and server DRAM into multi-year hyperscaler contracts and crowding out commodity modules.