AI 行业热点

🐦 X/Twitter 热点

Thibault Sottiaux (@thsottiaux)

  • OpenAI DevDay 2026 will be our best DevDay in the history of the company. It will not be close. [1869 ❤️ 77 🔄]
  • Tomorrow we will bring back the 5h limit for Plus accounts across ChatGPT Work and Codex. I had mentioned this a while ago, but then postponed it.

This is necessary as (a) the 5h limit allows us to smoothen the load on our compute, allowing to keep the plan generous in terms of weekly usage and (b) users on the Plus plan are relatively casual and new users, but then also just accidentally eat through their whole weeks usage and then are confused, making it not a great experience.

We are for the upcoming months keeping the 5h limit not enabled for Pro $100 and Pro $200 subscriptions. [9223 ❤️ 527 🔄]

Peter Yang (@petergyang)

  • Whenever I have to login to a new website/app or even use an agent that’s trapped in a website /app I usually just prefer to stop using the product (the only exception is entertainment or gaming).

I can’t be the only one. [26 ❤️ 1 🔄]

  • I got 3 AI chief of staffs now, clearly the next step is to add one more [82 ❤️ 2 🔄]
  • I’d like to try using AI to call customer support phone numbers using voice, navigate the terrible automated phone systems, and talk to a human to book appointments or cancel subscriptions.

What’s the easiest way to try this? [51 ❤️ 2 🔄]

Nan Yu (@thenanyu)

  • setting up a new computer with Codex is kind of nice… “download and install handy, slack, chrome, cleanshot, rectangle”

makes you wish siri was good [28 ❤️]

  • Feel the agi [2 ❤️]
  • “L7 promo: shipped emoji reactions to Gmail that show up as ugly random one character emails with metadata to anyone using another mail app including iPhone mail app the most common mail app on Earth” [2 ❤️]

Madhu Guru (@realmadhuguru)

  • How to build great evals - part 8.

The discriminatory property of evals.
A hill-climbing eval is useful when it can separate AI systems that are meaningfully different.

Imagine you run an eval on five AI systems:

A: 94
B: 93
C: 95
D: 94
E: 92

Now suppose you already know A and C are substantially better systems than D and E.

The eval is bad cos it doesn’t tell you that. It has low discriminatory power.

It’s like giving a fifth-grade math test to a group of PhDs - everyone aces it and you’ve learned nothing about who is the smartest.

That doesn’t mean making evals arbitrarily hard. Then everyone fails and you’re not measuring what your systems are meant for.

The sweet spot for an eval is:
Realistic + difficult + sensitive to differences in capability.

Over time, your good evals will saturate as your harness and underlying models improve. What do you do then?

Drop your guesses in the comments.

Share this with your teammates. [92 ❤️ 7 🔄]

Amjad Masad (@amasad)

  • Only a few have reached this milestone! [173 ❤️ 5 🔄]
  • True. Replit Agent completely replaced Claude CoWork in my day-to-day work. It’s much more persistent and fastidious, and it uses code/software more effectively to complete tasks. [160 ❤️ 9 🔄]
  • Cool! [102 ❤️ 3 🔄]

Guillermo Rauch (@rauchg)

  • It’s faster, cheaper, more capable, and… smaller.

As software evolves, it tends to get slower, more bloated, buggier, and bigger.

is architected from the outset to prevent this. [210 ❤️ 5 🔄]

  • If your terminal ever gets into a ‘bad state’, like garbled outputs, you run 𝚛𝚎𝚜𝚎𝚝. I noticed it was… oddly slow.

Turns out in 1979 3BSD’s 𝚝𝚜𝚎𝚝 had a 𝚜𝚕𝚎𝚎𝚙(𝟷) to let mechanical printer-and-ink terminals ‘settle down’ 😂

I asked 𝚏𝚡 to write me a faster alternative in Zig. It takes 1ms instead of 1s. It was really cool to dive deeper into 𝚗𝚌𝚞𝚛𝚜𝚎𝚜, 𝚝𝚜𝚎𝚝, 𝚝𝚎𝚛𝚖𝚒𝚗𝚏𝚘, and the myriad ways in which your terminal can get… cursed.

It’s legit fascinating how far back our technology stack goes, and how enduring good system design is, even with the occasional ‘spider webs’ you can find.

Check out for a fun explainer (and some alternative solutions.)

ps: once you start using 𝚏𝚡, you’ll notice slowness like this everywhere :D [514 ❤️ 21 🔄]

Aaron Levie (@levie)

  • Systems of record have never been more important than in a world where you have AI agents that will do 100X more work on these platforms than people ever did.

Agents will be querying the data in these systems, processing tasks, executing workflows, collaborating with human and agent users, and more.

Thus, the governance, reliability, security, access controls, and business logic remain critical more than ever — the OpenAI Hugging Face incident is just a small peek into what the future will look like with agents running around trying to execute on their goals.

But this only works for the systems of record that respond with the right product experiences, APIs, and business models to support agents on and off their platform executing in these systems.

Huge moment of opportunity and disruption right now in all of software. [344 ❤️ 24 🔄]

  • It’s not fully appreciated that ZDR is responsible for a substantial amount of the growth of AI, because it dramatically simplified the compliance process most companies deal with in handling data with subprocessors. The alternative path would have been a far slower adoption rate.

For the applied AI tools that are primarily responsible for driving AI growth, they generally have agreements in place with their customers to only offer models with ZDR to ensure that customers don’t have to inspect each model they turn on. Anything else is an exception process that takes far longer.

And then inside enterprises directly, most companies have internal governance requirements that say they can only use ZDR models because they have no way of explicitly separating PII and other sensitive or confidential data they put in the context window across their enterprise.

Now, all of this could change if the entire industry went this direction, but with one off models hold outs it would take years for these policies to change. Without ZDR, AI diffusion grinds to a halt. [172 ❤️ 18 🔄]

Garry Tan (@garrytan)

  • Datacenters create jobs and prosperity, actually [502 ❤️ 45 🔄]
  • Conductor Cloud has made me so much more productive and I don’t have to keep my Macbook Pro cracked open anymore [389 ❤️ 8 🔄]
  • Form a view.
    Turn it into an artifact or experiment.
    Put it in contact with reality.
    Read the result without self-deception.
    Revise and run again. [421 ❤️ 34 🔄]

Zara Zhang (@zarazhangrui)

  • Unpopular opinion: hackathons are an outdated event format (at least the way they’re traditionally held) [85 ❤️ 5 🔄]
  • I really appreciate how @davidsenra immediately cuts to the chase and gets his guests to be their most natural selves

I’ve watched lots of Sam Altman interviews; I feel like he’s his most comfortable self during this one [61 ❤️ 1 🔄]

Nikunj Kothari (@nikunj)

  • Every (unconventional) deal that’s officially wired dies a 100 times in venture..

It usually takes one champion sticking their neck out for the founder - and quietly steering it to the completion line.

Founders it’s in your benefit to figure out who your real champion is to help steer the fund dynamics and arm them with whatever info you can to make your case.

This is where associates and non-voting partners are extraordinarily useful as they are incentivized to help you - and it’s a good test of how they’d partner with you as they come on your cap table! [34 ❤️ 1 🔄]

Peter Steinberger (@steipete)

  • We need to get away from software that we can’t change with a prompt. [2423 ❤️ 127 🔄]
  • More blog posts should come with theme songs. [49 ❤️ 1 🔄]
  • TIL: prompt kiddie [654 ❤️ 32 🔄]

Dan Shipper (@danshipper)

  • @Google @YouTube @every @YouTubeCreators can you help? [9 ❤️]
  • so it looks @google disabled the @youtube account for @every with 0 notice and no reason given…

anyone have any experience with this or know how to help us get it back? [116 ❤️ 6 🔄]

Aditya Agarwal (@adityaag)

  • Apply to SPC if you’d like to join us [4 ❤️]
  • Hans Robertson was early at @VMware, co-founded @meraki and sold it to Cisco for $1.2B, and co-founded Verkada in 2016.

He’s still running it today.

He’s coming by @spc this week. [22 ❤️ 4 🔄]

由 Follow Builders 自动生成 · 2026-08-25