SGLang became the infra for serving LLMs. @PhAILabs is building the infra for AI-driven science.
Proud to see PhAI support the @sgl_project community at SGLang Summit — open infrastructure is how both AI and science scale.
📅 Nov 12–13 · Fort Mason, SF
Building AI products and watching markets share one annoying thing: the most useful data lives behind interfaces with no API.
The Sai API demo is a clean illustration of the fix. An app built on it is asked to sign in to Product Hunt through Google. Its own Windows cloud computer does it, live on screen, then returns today's top 5 launches with a screenshot.
What that pattern covers:
- desktop apps, old CRMs and admin panels nobody will ever build an API for
- sites that need a real login instead of a scraper
- routine checks you want done the same way every day, which Sai learns once and reuses
- several of these at once, since every agent gets its own machine
And the guardrail that would make me comfortable pointing it at anything sensitive: it pauses for approval before risky actions.
#1 on OSWorld 2.0 in partial success rate (79.38%). Works with Claude Code, Codex and Cursor.
https://t.co/NSPKcI1l1C
turning "trust me bro, I won’t sell" into something verifiable onchain
team allocations, investor vesting, community incentives, treasury reserves — anywhere you need a promise to actually mean something
and if the locked token can still earn dividends?
even better
Most productivity tools can tell you what needs to be done.
The harder question is:
What actually deserves your attention next?
A project rarely moves in a straight line.
You make a decision.
Reject an approach.
Discover something new.
Leave a problem unresolved.
Start working on something else.
Then come back days later.
By the time you ask, What should I do next?, the answer depends on everything that happened before.
That’s why a simple task list can only take you so far.
It knows what’s open.
It doesn’t always know why it matters now.
As AI makes it easier to research, write, build, and generate options, the bottleneck increasingly shifts toward deciding what deserves attention.
That’s the idea behind Next Move in https://t.co/plHxN8JuA0
Instead of simply giving you more things to do, it looks at the context around your work and surfaces a possible next action.
You stay in control. Jarvix suggests. You decide what happens next. Because personal AI shouldn’t just help you do more.
It should help you recognize what matters next.
curious about the approve or reject interface for Sai. if you spin up a computer fleet to copy cells, does it pause for confirmation before saving? #SaiFleet works as a quiet robosecretary.
See u tmr to talk about tokenized equities.
specifically how stock memes like base:0xb200000000000000000000cfbdf64a8706a94a01 are bootstrapping the initial liquidity of serious onchain infra.
I don't like to promise financial returns and it's also what's advised against by my general counsel and lawyers.
But @o1_exchange alpha listing is born different. it's for the long-term projects who can go thru liquidity cycles.
o1 alpha is like @binance alpha listing but for @coinbase@CoinbaseMarkets@base. Note that it has no ties to Coinbase / Base, but it's actively managed and researched by the alpha onchain analysts and the biggest @base traders.
NFA. DYOR.
All I want to say is that we're here tgt in this for the long term and we're winning it tgt on @base.
o1 alpha listing will be another listing similar like Coinbase / Binance / Upbit for the wealth effect, but much earlier stage and much longer term.
Sky is the limit.