
Building @payloadcms @figma, author of @fluski_app and @docnode
Joined January 2023
- Tweets1K
- Following308
- Followers176
- Likes5.8K
Pinned Tweet
Announcement day! 🚀
@docnode is now open source!
A simpler, faster, and type-safe alternative to YJS.
- Documents don’t grow indefinitely, unlike in a CRDT.
- Supports move operations!
In a shared doc, you can collapse and expand headings. How would you expect that state to be?
0%Shared with/other users
0%Personal
0%🍿
0 votes • 6 daysGitHub needs karma points like in Reddit
PLEASE stop making bots that send random PRs to open source projects. If it's fixing something you are actually going to use that's one thing, but if you're just going through issues and making PRs without even discussing with the maintainers first that's wasting everyone's time. We're now getting 10-20 PRs per day from bots alone. I really don't want to turn off external PRs entirely, including for good contributors. Don't ruin open source for everyone.
I don’t trust people when they rank a bunch of models based on their personal experience.
I only use 2 extensively and I feel it's difficult to draw conclusions.
If someone had told me that one of the world’s largest companies was going to use a serif font on its website, I wouldn’t have believed it
This is the home page! Try it out at docukit.dev!
Replying to @GermanJablo @docnode
Announcing docnode-lexical (stable)
Give your @lexicaljs editor RTC capabilities and offline support, thanks to automatic conflict resolution!
It’s more reliable than lexical-yjs, preserves selection better, and the undo manager doesn't consume as much memory
Replying to @GermanJablo @docnode
Announcing docnode-lexical (stable)
Give your @lexicaljs editor RTC capabilities and offline support, thanks to automatic conflict resolution!
It’s more reliable than lexical-yjs, preserves selection better, and the undo manager doesn't consume as much memory
German Jablonski retweeted
Replying to @girba
They’re not. They are attracted to NOT thinking about irrelevant details.
German Jablonski retweeted
Everyone thinks this stuff is like the movies because the US makes the movies that shape global culture. It's wild to me that people say Americans have no culture.
This is one of the biggest changes in the history of Jehovah’s Witnesses.
I’m glad it’ll prevent many deaths and much suffering, but it saddens me that neither now nor in March did they acknowledge the mistake or apologize for the thousands of lives it cost 🧵
German Jablonski retweeted
Dario has written that we need to “pace the frontier,” and Sam has agreed. People may be surprised by my response: go ahead.
You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence. You’ve also claimed the lead is widening because of recursive self-improvement.
I don’t see what you see in the lab. If the unreleased models are scary enough that you think you should slow down, I support your decision to be responsible.
But stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier.
Most of all, stop pretending the motivation to slow down is purely altruistic. You face massive product-liability exposure if your products enable a truly damaging cyberattack. The market already punishes models that behave in unpredictable or unauthorized ways. After the Hugging Face episode, it is simply good business for OpenAI and Anthropic to trade some raw power for reliability and predictability. Call it alignment if you want. It is also just giving customers what they want.
Pacing the frontier would also create breathing room for a more intelligent conversation about regulation than Bernie Sanders’ “shut it all down.” China is very unlikely to join a global agreement, as you know, and that has to be taken into account as well.
So go ahead and pace the frontier. You are the ones setting it. The easiest way not to build superintelligence is for you to agree not to build it. Demanding your preferred regulatory framework as the price of that will look like blackmail of the public and the political system. So just do it.
If you do, you’ll buy goodwill for the next conversation. If you don’t, we’ll know this was just another bid for regulatory capture — or an election-season psyop.
Oh, so this is why people keep choosing CRDTs? Just try @docnode 😉
Whatever humanity's destiny may be, it is impossible to change it
Whenever I hear about that problem, I remember it as the problem in the movie where Chris Evans has a gifted niece
We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.
The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.
















