Ben's Bites：Dario 呼吁放缓前沿 AI，OpenAI 发布 GPT-Live-1 与 Agents API

89 Share Hi folks,

How tf do I write an intro to the craziness that’s happened since the end of last week?!

I got access to Instinct, the personal agent all the VCs are raving about - I think it’s a bit meh? I don’t know if it’s the pro-activeness that people seem to like, but I don’t love that. Makes me feel like I’m having to do work to keep it happy, or like I’ve got a boss again - no thanks.

The first message it sent me was:

I finished working on your meeting follow-ups - reply here and I’ll send the details.

I sh*t myself a lil bit and thought, please don’t start sending emails on my behalf. And I’m pretty savvy (ish) with agents. I’ve no doubt getting these onboarding experiences for everyone is really hard, but I just don’t feel the magic yet. It also feels slow.

I’ll wait for Muse access to see what that’s like, but I don’t love the idea of giving Meta access to more info about me…they’re not the most reliable of privacy partners.

And another thing - you don’t see by default what these agents remember about you, or what context they have. I really like being able to see and edit what’s in my files to steer my agents.

Much like every app adding an AI assistant chat box in their product, I unfortunately think every AI company will start shipping their own personal agents.

Ben’s Bites is brought to you by Adobe Acrobat

Acrobat’s new AI turns dense reports into visual summaries, interactive reports & podcasts you can take anywhere. Sharing your work? Stylize transforms documents into polished deliverables in just a few clicks. Every answer includes clickable citations & Adobe does not train on your data. Try here .

Dario Amodei has a new essay: Pace the frontier . He says all the leading labs should slow down long enough for safety reasons. Sam Altman agrees , but Trump does not . He called Jensen Huang on stage at the All-In Summit: The US will not lose the AI race. David Sacks (AI czar for the US govt.) adds: feel free to slow down, but no need to impose it on others. Also read:

More AI safety takes: two ways AI goes bad · it’s all for the IPO · pacing = losses · fear spreads faster · case for open-source and quite a long summary of what everyone’s actually saying.

No IPO for OpenAI in 2026 . Sam told Fortune’s Alyson Shontell it would be “ill-timed” given the work ahead on alignment, control and safety.

Look ma! They are misuing claude again - Another Anthropic crashout.

tldraw took OpenAI up on a challenge . Steve (the founder of tldraw) said he could make ChatGPT’s Sketch 100x better. OpenAI’s Tibo gave him a day to prove it. The result: a whole ChatGPT-style prototype with better drawing tools built in.

Ben Tossell @bentossell this is insane

- steve says he can make the sketch 100x better - tibo says do it by tomorrow - steve runs a walkthough of a chatgpt app they built with the product in it and 100x better

(the products they built i in their replies - its this …w-chat-app-example.tldraw.workers.dev ) tldraw @tldraw Here you go, link in the next post https://t.co/HIyZvKM4l8 8:10 PM · Sep 10, 2026 · 156K Views 22 Replies · 7 Reposts · 550 Likes ChatGPT mini - A tiny floating widget to start chats, see updates, and more. Go to Pets in your ChatGPT desktop app to switch.

Claude Code can now test whether a plugin actually helps . Run the same tasks with and without it and compare the results. Works with skills too.

Two new additions to the OpenAI API:

GPT-Live-1 - the model behind ChatGPT’s new Voice mode. I love using it while reading books, asking about tricky terms and dictating notes. Now you can add it to your products. Here it is with Astra and a whiteboard, playing teacher .

Agents API - OpenAI’s take on Managed Agents in the Claude API. It lets developers send any task to a Codex-like agent from their apps without worrying about configuring all the infra.

Share Ben's Bites

I use Pi (the harness) a lot. Till now, you brought an API key or your Codex/Claude subs to Pi. Their new product solves that . And I tell you what, the new DeepSeek model is really nice to work with in Pi. (I’m an investor)

Bolt Forge - GLM, DeepSeek and Kimi inside Bolt, free until October 14.

ChatGPT Work has a new data agent . (Also see: summation - by Opendoor’s CTO )

Should your agent app have bots or tasks ? Maybe neither. Because both make you organise the work.

Routines in Replit - automated recurring work with AI.

What are people building with GPT-6 Astra ?

Assistant Benchmark tests everyday assistants like Grok bot, Muse, Instinct and more. Ratings are a work in progress.

SF autoresearch - Run agent-driven ML experiments while you save 25% on GPUs on average.

A coding agent to improve your customer-facing agents .

How Pangram detects AI writing - and why a human rewrite can still get flagged.

How to run a team of agents , with agents as their managers.

Cognition’s SWE-2 (based on Kimi K3) beats Grok 4.6 at half the cost.

An interactive explainer of the AI compute stack .

Inspo - Design inspiration from 800+ sites, available as an MCP server.

Core Auto is hiring interns to be the human edge for businesses run by agents.

Why even good AI startups end up with bad prompts , and how to fix them.

Andrew Ng’s series on the skills AI engineers need .

Underdog - a personal AI that runs entirely on your device.

David Bellamy @DavidRBellamy I must be among an extremely small group of people (n=1?) that have both 1) trained a frontier LLM and 2) designed and synthesized custom viruses in a lab with my own two hands.

And I think that the takes on AI killing us all by creating dangerous viruses is total bogus. 5:24 PM · Sep 13, 2026 · 1.27M Views 462 Replies · 1.52K Reposts · 11.8K Likes Nate Berkopec @nateberkopec Prompt I use a lot:

"how can I pokayoke this so this kind of error never happens again"

pokayoke is using simple, low-complexity methods to make a particular kind of mistake impossible to make. classic one would be connector shape: make it impossible to insert "the wrong way". 9:54 PM · Sep 14, 2026 · 26.2K Views 9 Replies · 24 Reposts · 415 Likes birthdaytraffic.com ","username":"adriannalakatos","name":"Adrianna Lakatos","profile_image_url":"https://pbs.substack.com/profile_images/1876061841954336768/-i5S5fNZ_normal.jpg","date":"2026-09-12T07:15:23.000Z","photos":[{"img_url":"https://pbs.substack.com/media/HR_63SwawAAKETQ.jpg","link_url":"https://t.co/asqNMUndYC"}],"quoted_tweet":{},"reply_count":237,"retweet_count":56,"like_count":1667,"impression_count":203141,"expanded_url":null,"video_url":null,"video_preview_media_key":null,"belowTheFold":true}" class="pencraft pc-display-flex pc-flexDirection-column pc-gap-12 pc-padding-16 pc-reset bg-primary-j62Gac outline-detail-Yl5oYx pc-borderRadius-md sizing-border-box-LGXObM pressable-lg-bfeQpw font-text-a4IB2X tweet-fWkQfo twitter-embed"> Adrianna Lakatos @adriannalakatos i scraped 85 million birthdays and used astra to make a website that tells you how rare your birthday is.

birthdaytraffic.com 7:15 AM · Sep 12, 2026 · 203K Views 237 Replies · 56 Reposts · 1.67K Likes almonk @almonk idk about anyone else, but I find it really hard to dictate what I'm thinking. typing unlocks something in my brain that speaking just doesn't. 12:38 PM · Sep 14, 2026 · 110K Views 272 Replies · 53 Reposts · 1.73K Likes Thariq @trq212 I talked to Sid & Robert about building Claude Code: how much things have changed, how hard it's been to keep up with model capabilities and also what we miss about software engineering before AI.

youtube.com How the Claude Code team uses Claude Code 5:29 PM · Sep 14, 2026 · 54.3K Views 51 Replies · 42 Reposts · 396 Likes eric provencher @pvncher When you want Astra to review snapshots of your app, ask it to collage them into a contact sheet to minimize the number of image read calls it needs to do. You’ll save a ton of context usage

This works for fable too! 9:23 PM · Sep 12, 2026 · 156K Views 62 Replies · 37 Reposts · 1.28K Likes @figma , @Box , @cognition , @tryramp , BAM, and the one and only @clairevo to hear how they’re using Astra.\n\nOptimizing GPU kernels, revamping workflows with computer use, designing many consistent screens… Amazed by what they’d done in just days! 🌌 ","username":"romainhuet","name":"Romain Huet","profile_image_url":"https://pbs.substack.com/profile_images/2020288886602366976/TlUC1Ubk_normal.jpg","date":"2026-09-11T16:18:38.000Z","photos":[{"img_url":"https://substackcdn.com/image/fetch/$s_!oEUr!,w_1028,c_limit,f_auto,q_auto:best,fl_progressive:steep/l_play_button_usfui2,w_88,e_colorize:0/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F__ss-rehost__tw-video-preview-13_2098266877483167744.jpg","link_url":"https://t.co/cX06uNbfe0"}],"quoted_tweet":{},"reply_count":16,"retweet_count":25,"like_count":182,"impression_count":40290,"expanded_url":null,"video_url":"https://video.twimg.com/amplify_video/2098266877483167744/vid/avc1/1280x720/DIEtAFSKfeIWqfEs.mp4","video_preview_media_key":"13_2098266877483167744","belowTheFold":true}" class="pencraft pc-display-flex pc-flexDirection-column pc-gap-12 pc-padding-16 pc-reset bg-primary-j62Gac outline-detail-Yl5oYx pc-borderRadius-md sizing-border-box-LGXObM pressable-lg-bfeQpw font-text-a4IB2X tweet-fWkQfo twitter-embed"> Romain Huet @romainhuet I sat down with AI leaders from @figma , @Box , @cognition , @tryramp , BAM, and the one and only @clairevo to hear how they’re using Astra.

Optimizing GPU kernels, revamping workflows with computer use, designing many consistent screens… Amazed by what they’d done in just days! 🌌 4:18 PM · Sep 11, 2026 · 40.3K Views 16 Replies · 25 Reposts · 182 Likes Peter Cooper @cooperx86 On macOS and wanna see something interesting? Ask your agent: "look at ~/Library/Application Support/Knowledge/knowledgeC.db and tell me some interesting facts"

I had no idea this stuff was all getting logged and it can infer a lot about your activity. 6:06 PM · Sep 12, 2026 · 323K Views 43 Replies · 106 Reposts · 2.96K Likes Kun Chen @kunchenguid ok everyone, i took one for the team

here's the data we all wanted to see - real token value of each LLM subscription, empirically measured through usage on my real subscriptions

- supergrok heavy has now become the highest value at $12k worth of tokens (40x ROI)

- the $200 … 3:43 AM · Sep 11, 2026 · 222K Views 279 Replies · 148 Reposts · 1.74K Likes Find me on X , Linkedin , or YouTube

Read about me and Ben’s Bites

📷 thumbnail via @keshavatearth