Filippo (@neysofu) header
Filippo (@neysofu) profile picture

Bits mangler. Cera

Concent of Saunt Tredegarh
Joined May 2016
Photo by Filippo (@neysofu)Photo by Filippo (@neysofu)Photo by Filippo (@neysofu)Photo by Filippo (@neysofu)Photo by Filippo (@neysofu)Photo by Filippo (@neysofu)Photo by Filippo (@neysofu)Photo by Filippo (@neysofu)Photo by Filippo (@neysofu)Photo by Filippo (@neysofu)
Constructive Specifications: A Short Story blog.neysofu.me/constructive…
2
2
2
753
Filippo retweeted
Without getting into too much detail about the conversations I’ve had with my friends at Anthropic, it’s become obvious to me that their long-term ambition is to continue running these small-to-large experiments while scaling experiment throughput with AI agents. Now, researchers already use smaller models to test architectures and optimizers. However, what I’m hearing is that they want to do this on a much bigger scale. Think millions of smaller, cheaper experiments with models that have a billion parameters, with the most promising discoveries being validated at 10 billion, then 100 billion, and so on and so forth. The idea is to have autonomous research agents explore this enormous hypothesis space and identify which architectures and training methods scale best. Now, this probably sounds obvious to a lot of people already in the AI field. However, from what I’m hearing, there’s a lot of success happening right now, and they’re continuing to scale up this experimentation, which is quite exciting to hear.
can tell from the latest haiku :)
23
20
2
640
33,225
neon + t3.codes + boat.dev is the stuff of dreams
1
32
we married on the 11/11 and @calcom just hit $11.11M in ARR 🔥
133
13
3
1,465
95,381
Filippo retweeted
>engineer says an agent swarm secured his software >ask if it’s formal methods or 8 million lines of smoke tests >he doesn’t understand >pull out illustrated diagram explaining formal methods and smoke tests >he laughs and says “it's good security sir” >check >it smoke tests
12
5
1
249
10,250
Composability really is the name of the game. We need to build a culture and ecosystem of open specs, as opposed to monolithic verification efforts.
The bull case for formal verification is "Here is a reusable library of properties/specs you can import in your code" not "Here is a few million LOC of Lean for you to go over, but the model passes!"
2
6
2
45
2,853
Filippo retweeted
The language matters more with agents, not less. The compiler is the only reviewer that reads every line
3
15
1
73
3,004
Filippo retweeted
💭 unpopular opinion: type theories of Lean / Rocq are not complex and heavyweight enough. we need low-level, type-checked control over runtime *and* programs and proofs in the same lang maybe dependently typed lang with OxCaml's modes. even better: graded modal dependent types!
Lean, Rocq, etc. seem very complex and heavyweight for many applications but they serve as least common denominators, and now that AI can write in type theories, why not? Still, it feels like a smell and I sense something different is around the corner, not sure what though
23
17
1
194
9,485
The bull case for formal verification is "Here is a reusable library of properties/specs you can import in your code" not "Here is a few million LOC of Lean for you to go over, but the model passes!"
32
22
11
305
26,128
prediction: victor will find and showcase a decent intelligence primitive within the next year, that's not matmul
I'm (ofc) super interested in studying systems that do NOT have attention anywhere (nor anything like it), yet that show signs of useful intelligence (like answering English questions). I'm collecting a list of these. Appreciate pointers. Could be anything
85
All your opinions are wrong
This quoted post is unavailable.
57
That's right! Audio spectrography goes in the transformer hole.
So Astra is able to identify sounds from mel spectrograms zero-shot. I don't think we've scratched the surface of what this model can do (and this is light reasoning btw)
21
146
11
3,479
111,697
Addicted to the feeling of finishing my day tired and content, knowing I did good
84
JSON5 should have won the config format war instead
I’m having a real problem: why is TOML a thing? Someone decided to reinvent INI files from the 1980s and discard all of the reasons we moved on from them over the last 40 years Why would I choose TOML over JSON, XML, or even YAML?
1
118
I want ai math posts back on my timeline, tyvm
1
117
Everybody's solving open problems with AI, but nobody's thinking beyond the context size You can't one-shot UGC or Riemann. You need a harness with agents operating on a graph of theorems, lemmas, proofs, representations
79
brb buying a military-grade intra-oral mic to talk to Codex on the subway
3
115
Replying to @SeanSankerJr
Current programming languages were designed with human ergonomics in mind. Humans do not write code anymore. So we should throw away all old languages and design new ones, geared toward the machines to write code in.
13
6
2
73
5,377
5 years ago
I'm not sure how useful this thing will actually be in my day to day programming activities, but A for effort. #Rust #GitHubCopilot
1
90
It's hard to overstate the amount of damage caused by the original quote It made us all believe that having someone to blame is more important than improving outcomes
2
1
323