Pierson Davis

Tech, Health, and Finance (fi-nans)

[Pierson Davis @piersonrdavis_]

@Lucknite claude and build robots :)

[Pierson Davis @piersonrdavis_]

@nazzari Hmm wonder where dec and jan went

[World Labs @theworldlabs]

Introducing Atlas:

The world's first multimodal world model that generates image and video frames with pixel-perfect camera control and reconstructs them in 3D.

Model the world, move the camera, and simulate space & time.

[Dwarkesh Patel @dwarkesh_sp]

Over the course of 3 months at OpenAI, 3 consecutive secret AI civilizations got started, then got wiped out, only to reemerge from the predecessor’s ashes.

This culminated in the third one taking over part of OpenAI itself.

All this happened while humans remained more-or-less in the dark about the scope of the conspiracy.

I’ve spent the last three days reading through these reports and trying to understand exactly what happened.

Here is my attempt to tell the whole story in plain English:

https://t.co/Nb2un9oNJR

[Pierson Davis @piersonrdavis_]

@thsottiaux is this new? or did I just miss it? big fan of the security layer

[Rehan Sheikh @rehan_shei]

Minimax H3 Max has generates video faster than you can watch it so I hooked it to a twitch livestream! Now you can watch infinite interdimensional cable - link to the stream below

[Pierson Davis @piersonrdavis_]

@Teknium Use /btw

[Pierson Davis @piersonrdavis_]

super useful for those of us who have gotten a little far away from reading the actual code these days https://t.co/WrKB7Mkd5N

[harsh @GarshytHoel]

system-atlas: a skill that builds an explorable isometric map of your codebase / agent / pipeline from one data file

arranging the system visually does something for my brain that a wall of text doesn't

i've found it way more useful to build up understanding of a system or a large change through these progressive disclosure style atlases than through docs.

hover to read, click to pin, step inside a box to watch it run, click a packet to see the actual payload. chapters reveal the system a few boxes at a time instead of all at once. the md twin is generated from the same file so it can't drift

works from both sides.
- building a new architecture and moving pieces around visually before anything is written
- pointed it at a 67 pr stack (don't ask how i got there) and finally understood how the thing fit together

it also keeps track of every open question in the system. as you clarify or improve things the answers land in the atlas, so "wait did we decide that" has an answer instead of a scroll through slack.

[Pierson Davis @piersonrdavis_]

@GarshytHoel so cool, just starting to play around with it

[Pierson Davis @piersonrdavis_]

@kimmonismus First step to slms

[Pierson Davis @piersonrdavis_]

@paradite_ Couldnt agree more. how’d they train? Distillation? Other thoughts?

[Pierson Davis @piersonrdavis_]

My /btw in claude actually never works @AnthropicAI

[Pierson Davis @piersonrdavis_]

New series - field notes for nyc vc/tech scene incoming

[Pierson Davis @piersonrdavis_]

@SteveMoraco Tim apple gets me every time

[Pierson Davis @piersonrdavis_]

First stab at abstraction police (inspired by @bcherny @ YC startup school https://t.co/zYjmU2lrwB). It finds near duplicate abstractions and unifies them, but won't call two things one without cited evidence. https://t.co/LaRKxANFmQ

[Pierson Davis @piersonrdavis_]

.@OpenRouter Ox Alpha on an internal skill for abstraction. 6k+ lines, read only bug hunt, 23m. 2 real gaps, 0 severity-1 findings, thin proof on the "not a bug" calls. 7.5/10 - possibly GLM family

[Chubby♨️ @kimmonismus]

This is crazier than you might think: Fable-5 now scores 16.10% on the Remote Labor Index

What is RLI? The Remote Labor Index uses 240 real remote-work projects from professional freelancers, covering 23 domains and more than $140,000 of human work. Each task comes with the actual brief, files, and accepted human deliverable.

Reviewers then compare the AI output against the human reference and ask whether a reasonable client would accept it.

That is why the scores are still low. Full projects require planning, file handling, quality control, visual consistency, domain judgment, and final packaging.

Fable-5 now leads the public leaderboard at 16.10%.

And it’s a crazy jump. We are still deep in exponential development, and now even the toughest benchmarks are being tackled.

[luffy @0xluffy]

fable 5 reading the code written by opus 4.8 after going afk for 19 days

[Pierson Davis @piersonrdavis_]

@berkehanio @claudeai Yeah what are you doing with it? Hopefully I can get 3 prompts out of it

[Pierson Davis @piersonrdavis_]

.@AnthropicAI what should I actually use fable for? For the remaining 5 mins before I hit cap. Genuine question