AI 行业热点

🐦 X/Twitter 热点

Thibault Sottiaux (@thsottiaux)

  • This might have been my best day so far at oai. Ridiculous amounts of fun and intensity. Future is bright [4541 ❤️ 70 🔄]
  • The real story is this one [1858 ❤️ 58 🔄]
  • Sergio is cooking up magic with collaborative space right in ChatGPT. Improving leaps and bounds every day. [1736 ❤️ 25 🔄]

Peter Yang (@petergyang)

  • Learn more about Gemini 3.8 Live here:
  • I built an AI language learning app that teaches me Japanese via live voice calls.

Here’s my new tutorial that shows you exactly how I built this, including:

→ Using my spec skill to create key designs
→ Setting up Gemini Live APIs for voice
→ Creating diorama art with Nano Banana

You can follow along to build your own app to learn 100 phrases for any language before a trip.

📌 Watch now: [75 ❤️ 7 🔄]

Nan Yu (@thenanyu)

  • You can just fix things [36 ❤️ 2 🔄]

Madhu Guru (@realmadhuguru)

  • The most common mistake teams make is to treat evals as an additional QA step once the agent is built.

AI products are fundamentally different.

Your evals are your product spec. [68 ❤️ 4 🔄]

Thariq (@trq212)

  • if you’re getting into game design, there are so many good resources out there to learn from! [116 ❤️ 4 🔄]
  • a big advantage of doing planning like this is that it’s a lot more token efficient compared to raw HTML-

the model doesn’t need to remake the components or logic to do common things like state machines, diagrams, code snippets, etc. [898 ❤️ 29 🔄]

  • my favorite name for this pattern is “local hands”, Claude runs in the cloud but can access your files locally- also coming to cowork [828 ❤️ 16 🔄]

Amjad Masad (@amasad)

  • “Promote my app on tiktok” [75 ❤️ 6 🔄]
  • Really proud of how many family-run businesses are powered by Replit. [111 ❤️ 7 🔄]
  • The US is catching up on open-weights models. [278 ❤️ 20 🔄]

Guillermo Rauch (@rauchg)

  • Introducing gdp-ts: Ghosts of Departed Proofs for TypeScript.

gdp-ts is a library, linter and AI skill for safer API design. Under this contract, sensitive functions require ‘proofs’ that the caller performed an authorization check.

The typechecker verifies these proofs at compile time, preventing your team and agents from shipping catastrophic security (and other kinds of) bugs.

While these patterns have existed for quite some time, especially in ecosystems like Haskell, ① human code review and ② cognitive and syntactic overhead made these solutions niche.

The situation is now inverted. Agents are writing more code than we can review, and they thrive in tight loops with hard constraints that would frustrate us. We see this with the rise of Rust, borrow checker, code aesthetics debate and all.

The README and examples model a real-world Vercel API product constraint: changing the password on a Project requires a proof of a certain role + a certain entitlement. Thanks to Matt Noonan and Ollie Charles for their research in this space. [2176 ❤️ 107 🔄]

Aaron Levie (@levie)

  • A decent amount of agent adoption and deployment is bottlenecked by being able to test, tune, and optimize agents on “real” work environments. The files agents need to access, the CRM environments, email, and more.

It’s impossible to know how agents are performing if you don’t understand how well they’re executing on your evals and how they’ll perform in the future with a model change or workflow upgrade. Every enterprise has to do this 1 by 1 right now, which is obviously slow and tedious.

Every enterprise is going to have someone managing evals, infrastructure for building and running evals and simulated environments, and more. Huge space. [143 ❤️ 11 🔄]

Ryo Lu (@ryolu_)

  • remembering Steve

i grew up with beige boxes. then i moved to Canada and saw the iMac G4.

how can computers be so beautiful?

i couldn’t afford a Mac, so i made my PC look like one. changed the icons, the fonts, the windows. learned to make those shiny, lickable buttons. that’s how i got into interface design. i saw something i loved and wanted to make things like it.

i burned DVDs of Steve’s keynotes and studied them. watched them again and again. there was so much care in the products, and so much joy in showing people what they could do.

my first iPod was a black nano. i carried it everywhere. then the iPhone happened and changed the world.

it felt like anything was possible.

since he died, i’ve felt us getting stuck in a loop. constant refinement. better supply chains. more ways to keep people scrolling. so much talent spent on tiny improvements, so little room for something truly great and heartfelt.

the keynotes get more cinematic every year. i still miss the live demos. someone onstage playing with something they helped make. the excitement, the jokes, the things that went wrong. it felt human. it was fun.

we made computers faster. phones thinner. added more cameras. made machines smarter than us at more and more things. somehow, we’ve also become more disconnected and lonely.

we spend so much energy trying to beat humans. i wish we spent more time learning from people’s lives and making them better.

when i work, i still sometimes think: what would Steve do?

i’ll never know. i’m sad i never got to work with him. but his work reached a kid with a beige PC and changed the course of his life. i wonder how many others have a story like mine.

i’ve kept these words with me:

“the people who are crazy enough to think they can change the world, are the ones who do.”

there’s still so much worth making. things we can put our hearts into. things that make someone feel what i felt when i first saw that iMac.

thank you, Steve. 💛 [566 ❤️ 27 🔄]

  • add ryOS Subtitles to Chrome [25 ❤️]
  • made a lil thing for language learners:
    watch Netflix with subtitles in any two languages, plus pronunciation guides for Japanese, Chinese, and Korean. customizable styles, hand-picked defaults.
    enjoy 🍿 [798 ❤️ 29 🔄]

Garry Tan (@garrytan)

  • AGI Science Loops are coming, and there will be Muse/Instinct for those AGI Science Loops. Halmos is building that. [481 ❤️ 33 🔄]
  • Harnesses from labs have an incentive to burn tokens, which means harnesses from startups have a real utility, like how Grep can automatically swap token burn into deterministic tested code that is repeatable just by observing agent use [512 ❤️ 41 🔄]
  • We should have competence and good outcomes but we can’t do that without leadership and accountability

High speed rail has been waylaid into a bulk transfer of taxpayer dollars into special interests with no good working outcome in sight [187 ❤️ 8 🔄]

Matt Turck (@mattturck)

  • The guy who thought he was too late to crypto investing [46 ❤️ 2 🔄]

Zara Zhang (@zarazhangrui)

  • Cool! [8 ❤️]

Nikunj Kothari (@nikunj)

  • [56 ❤️]

Dan Shipper (@danshipper)

  • what we promise @every:

as the frontier advances, so do you. [87 ❤️ 2 🔄]

  • waiting for your dot to pick up [29 ❤️]
  • great article by @ArielleShipper [18 ❤️]

Aditya Agarwal (@adityaag)

  • Final few spots left [7 ❤️ 1 🔄]
  • Three Stanford degrees.

Berkeley EECS faculty since 2016.

Cofounder of @physical_int.

@svlevine builds the algorithms that let robots learn from experience, linking what they see to how they move.

He’s coming to @spc tomorrow. [152 ❤️ 9 🔄]

  • It would be interesting to be able to run the Muse/Dot “computer” locally.

Feels like that is a better agent environment than increasingly locked down MacOS. [69 ❤️ 1 🔄]

由 Follow Builders 自动生成 · 2026-10-06