今日科技推特精选 - 2026-04-17

2026年4月17日科技每日简报

Today's top tech conversations are led by @swyx, whose post about 'RT @claudeai: Introducing Clau...' garnered the highest engagement. Key themes trending across the top stories include https, model, claude, today, openai. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. swyx (Group Score: 1268.8 | Individual: 58.2)

Cluster: 43 tweets | Engagement: 17188 (Avg: 154) | Type: Tech

RT @claudeai: Introducing Claude Opus 4.7, our most capable Opus model yet.

It handles long-running tasks with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back.

You can hand off your hardest work with less supervision. https://t.co/PtlRdpQcG5

See 42 related tweets

@Shipper_now: Claude Opus 4.7 was just released...

And now it is the default model in Shipper, the #1 AI app buil...

@bcherny: Opus 4.7 is in Claude Code today. It's more agentic, more precise, and a lot better at long-running ...
@svpino: Be careful with moving it all over Opus 4.7.

The new model uses more tokens and will eat up your s...

@omarsar0: What you need to know about Opus 4.7

Takes instructions literally
Better vision means improved ...

@MatthewBerman: Unexpected drop\n\nQT @claudeai: Introducing Claude Opus 4.7, our most capable Opus model yet.

It h...

2. garrytan (Group Score: 593.3 | Individual: 39.0)

Cluster: 23 tweets | Engagement: 1584 (Avg: 213) | Type: Tech

RT @perplexity_ai: Today we're releasing Personal Computer.

Personal Computer integrates with the Perplexity Mac App for secure orchestration across your local files, native apps, and browser.

We’re rolling this out to all Perplexity Max subscribers and everyone on the waitlist starting today. https://t.co/kxgFQFo7BB

See 22 related tweets

@cryptopunk7213: this is insanely impressive

perplexity just launched an AI employee that works for you 24/7 and run...

@MatthewBerman: I have to say something...please don't be mad.

OpenClaw has been nearly unusable for the past week...

@MatthewBerman: I’m going to be testing this. Perplexity is shipping at impressive speeds. But can they dethrone Ope...
@rohanpaul_ai: Perplexity just launched Personal Computer, a Mac feature that lets AI work across local files, nati...
@StockSavvyShay: Perplexity just launched Personal Computer which is a new tool in the $AAPL Mac app that securely wo...

3. ajambrosino (Group Score: 412.8 | Individual: 43.0)

Cluster: 15 tweets | Engagement: 1680 (Avg: 267) | Type: Tech

RT @OpenAI: Codex for (almost) everything.

It can now use apps on your Mac, connect to more of your tools, create images, learn from previous actions, remember how you like to work, and take on ongoing and repeatable tasks. https://t.co/UEEsYBDYfo

See 14 related tweets

@nvidia: Congratulations @OpenAI on bringing Codex to more of the software workflow. 🎉

Codex is becoming a s...

@Remotion: Check out the new Codex Plugins, you can find us there!

Within your Codex app you can now install ...

@minchoi: This is wild.

OpenAI is just turned Codex into a full AI operator.

Mac apps, images, automations, ...

@rohanpaul_ai: OpenAI just expanded Codex from a coding assistant into a desktop agent that can see, click, type, r...
@OpenAIDevs: RT @thsottiaux: Codex just got a lot more powerful.

Computer use, in-app browser, image generation ...

4. huggingface (Group Score: 383.7 | Individual: 49.2)

Cluster: 15 tweets | Engagement: 2404 (Avg: 122) | Type: Tech

RT @Alibaba_Qwen: ⚡ Meet Qwen3.6-35B-A3B：Now Open-Source！🚀🚀

A sparse MoE model, 35B total params, 3B active. Apache 2.0 license.

🔥 Agentic coding on par with models 10x its active size 📷 Strong multimodal perception and reasoning ability 🧠 Multimodal thinking + non-thinking modes

Efficient. Powerful. Versatile. Try it now👇

Blog：https://t.co/EXx5y466su Qwen Studio：https://t.co/bg4tAU1p74 HuggingFace：https://t.co/w4pDX14DZS ModelScope：https://t.co/SuRyLzdQiO API（‘Qwen3.6-Flash’ on Model Studio）：Coming soon～ Stay tuned

See 14 related tweets

@UnslothAI: Qwen3.6-35B-A3B can now be run locally!💜

The model is the strongest mid-sized LLM on nearly all ben...

@vllm_project: 🎉 Congrats @Alibaba_Qwen on the first open-weight Qwen3.6! Stronger agentic coding and a new thinkin...
@lmsysorg: 🚀 Congrats @Alibaba_Qwen on releasing Qwen3.6-35B-A3B — day-0 support is now live in SGLang!

The fi...

@Prince_Canuma: Coming to MLX 🚀\n\nQT @Alibaba_Qwen: ⚡ Meet Qwen3.6-35B-A3B：Now Open-Source！🚀🚀

A sparse MoE model, ...

@eliebakouch: yeah you know.... moe model are fundamentally limited... dense model are way better look at gemma4 a...

5. TheoBuildsAI (Group Score: 373.3 | Individual: 49.9)

Cluster: 12 tweets | Engagement: 1824 (Avg: 219) | Type: Tech

RT @rahulchhabra07: you can now control things with your brain. literally.

we're building the most wearable BCI on the planet, with @sabicap, backed by @khoslaventures @accel @initialized & @kevinweil.

we collected the world’s largest neural dataset and trained the most capable Brain Foundation Model.

then we invented a new class of biosensors powered by custom ASICs.

type without typing. click without clicking. a cap that lets your brain do the work.

we’re sabi.

See 11 related tweets

@rohanpaul_ai: This is the AlphaFold moment for brain-computer interfaces.

Sabi releases the Sabi Cap. The first ...

@Scobleizer: The thinking cap is coming.

I didn't even know about this one, heard about a different one.

Ther...

@Parul_Gautam7: Sabi came out of stealth today

-thoughts don't stay in your head anymore -non-invasive brain-comput...

@dr_cintas: This beanie can read your brain and control your devices 🤯

Sabi is building a non-invasive brain-co...

@svpino: I want to try this. Now!

I've read a ton about Brain Computer Interfaces, and while it sounds like ...

6. Mayhem4Markets (Group Score: 243.3 | Individual: 33.7)

Cluster: 12 tweets | Engagement: 22 (Avg: 191) | Type: Tech

The benchmarks for Opus-4.7 seem genuinely impressive.

A sizable step up from Opus-4.6 on agentic coding, tool use and financial analysis.

Will be running some tests on it soon to see how well it does! https://t.co/TfRROAd4RP\n\nQT @Mayhem4Markets: Big news!

Anthropic's Claude Opus-4.7 is going to officially drop anytime now.

Already seeing it on their website. 👀

Does this mean software stocks get nuked again or can we get a break? 😅 https://t.co/YzDtpULEr1

See 11 related tweets

@Mayhem4Markets: Claude Opus-4.7 is released and $IGV isn't nuked.

Are the SaaSpocalypse lows in now? 🤔 https://t.co...

@scaling01: for all the people calling Opus 4.7 a mid update lmao https://t.co/vG981kCojT\n\nQT @ValsAI: The new...
@kimmonismus: Opus 4.7 Benchmarks out! Very solid upgrade to Opus 4.6!

Compared to Opus 4.6:

-SWE Bench Pro +11%...

@Mayhem4Markets: Big news!

Anthropic's Claude Opus-4.7 is going to officially drop anytime now.

Already seeing it o...

@wallstengine: Anthropic launched Claude Opus 4.7 today across all Claude products and kept pricing unchanged from ...

7. Parul_Gautam7 (Group Score: 189.5 | Individual: 43.1)

Cluster: 6 tweets | Engagement: 514 (Avg: 88) | Type: Tech

RT @HeyGen: We built our launch video in Claude Code using HyperFrames.

Now it's yours.

Open source, agent-native framework. HTML to MP4.

$ npx skills add heygen-com/hyperframes

RT + Comment "HyperFrames" to get the full source code of this launch video (must follow) https://t.co/vsRtZ6gQsb

See 5 related tweets

@rohanpaul_ai: HeyGen just open-sourced HyperFrames, it lets AI agents turn HTML, CSS, and JavaScript into MP4, MOV...
@Origin_AI_01: HTML to MP4 from the terminal changes the game - when Claude Code becomes your video editor, creatio...
@Parul_Gautam7: HyperFrames turns video editing into a terminal-native workflow.

HTML → MP4 with one-shot prompts ...

@TheoBuildsAI: Video editing from the terminal wasn’t on my 2026 bingo card. But this + Claude Code might change ev...
@ai_explorer25: Open sourcing the exact framework behind your own launch video is a bold move\n\nQT @HeyGen: We buil...

8. Shipper_now (Group Score: 180.4 | Individual: 31.0)

Cluster: 8 tweets | Engagement: 50 (Avg: 24) | Type: Tech

🚨 BREAKING: Telegram-first vibe coding is here

Build any full Telegram bot by talking to Claude Opus 4.6 https://t.co/mkRyl9AW0x\n\nQT @durov: ⚡️ You can now create agentic bots in just 2 taps 🤖

Ask the devs of the services you use to support it 👇

https://t.co/8ICnEI2Avy https://t.co/oqKbsYMh7G

See 7 related tweets

@Shipper_now: 🚨 JUST IN: Build Telegram bots from prompts

The world's first-ever AI Telegram Bot Builder is here ...

@chhddavid: 🚨 BREAKING: Telegram-first vibe just dropped.

You can now build Telegrams bots while chatting to Cl...

@chddaniel: Claude in @shipper_now can build Telegram bots...

This is genuinely terrifying. https://t.co/DvZXSP...

@chhddavid: this is terrifying.\n\nQT @shipper_now: 🚨 BREAKING: Telegram-first vibe coding is here

Build any fu...

@chddaniel: this is scary.\n\nQT @shipper_now: 🚨 BREAKING: Telegram-first vibe coding is here

Build any full Te...

9. kimmonismus (Group Score: 180.0 | Individual: 36.5)

Cluster: 6 tweets | Engagement: 131 (Avg: 373) | Type: Tech

OpenAI released GPT-Rosalind today, a frontier reasoning model purpose-built for life sciences research, available in trusted-access preview to customers like Amgen, Moderna, and the Allen Institute.

On the benchmarks, it's competent rather than revolutionary, leading on BixBench, beating GPT-5.4 on 6 of 11 LABBench2 tasks, and cracking the 95th percentile of human experts on a curated RNA prediction task with Dyno Therapeutics.

But the real story isn't the scoreboard: The structural competitor here isn't Anthropic either; it's Isomorphic Labs (Google spinoff by Demis Hassabis), whose IsoDDE model from February was called "an AlphaFold 4" by researchers and is already embedded in billion-dollar deals with J&J, Lilly, and Novartis.

Rosalind is a reasoning and orchestration layer for scientific workflows, not a structural biology engine that folds proteins or designs molecules. Whats interesting is the pattern it confirms: frontier labs are fragmenting vertically, building industry-specific forks rather than scaling one horizontal generalist, and deploying them through trusted-access gates that look more like export controls than API keys. Read alongside Anthropic's Opus 4.7 and OpenAI's aggressive Codex super-app update on the same day, GPT-Rosalind is the less flashy release, but arguably the one that tells you most about where the business model is heading tbh.\n\nQT @OpenAI: Introducing GPT-Rosalind, our frontier reasoning model built to support research across biology, drug discovery, and translational medicine. https://t.co/PubLU0FkSv

See 5 related tweets

@Vtrivedy10: “specialized models/agents will totally get eaten by general purpose everything” 🙄

this is awesome!...

@rohanpaul_ai: OpenAI just launched GPT-Rosalind, a new purpose-built life sciences model series, built to help sci...
@WesRoth: OpenAI introduced GPT-Rosalind, a frontier reasoning model specifically architected for the life sci...
@cocktailpeanut: https://t.co/SThh33IQL9\n\nQT @OpenAI: Introducing GPT-Rosalind, our frontier reasoning model built ...
@techdaily24: OpenAI has launched GPT-Rosalind, a new frontier reasoning model specifically engineered for advance...

10. HessianFree (Group Score: 172.9 | Individual: 31.1)

Cluster: 7 tweets | Engagement: 100 (Avg: 73) | Type: Tech

> > anon asked for one more state 

> > we added zero 

> > +600 MB 

> > +5 benchmark points 

> > 75.5 avg at 1.75 GB 

> > still ~1/9 the size of Qwen3 8B 

> > shout out brahmagupta 

> > zero mattered\n\nQT @PrismML: Today we’re announcing Ternary Bonsai: Top intelligence at 1.58 bits

Using ternary weights {-1, 0, +1}, we built a family of models that are 9x smaller than their 16-bit counterparts while outperforming most models in their respective parameter classes on standard benchmarks.

We’re open-sourcing the models under the Apache 2.0 license in three sizes: 8B (1.75 GB), 4B (0.86 GB), and 1.7B (0.37 GB).

See 6 related tweets

@Prince_Canuma: Awesome release, congrats to the. @PrismML team!

It comes with day-0 support on MLX thanks to some ...

@HessianFree: RT @PrismML: Today we’re announcing Ternary Bonsai: Top intelligence at 1.58 bits

Using ternary wei...

@adrgrondin: New models from @PrismML.

Incredible intelligence density.

The new models outperforms other models...

@Scobleizer: The models are getting smaller. Great for OpenClaws and Hermes. Gotta heat them up!

Yesterday some...

@HessianFree: RT @TeksEdge: 🌳 PrismML drops Ternary Bonsai — top intelligence at 1.58 bits!

This is exciting!

Te...

11. nummanali (Group Score: 172.5 | Individual: 61.0)

Cluster: 4 tweets | Engagement: 5261 (Avg: 244) | Type: Tech

Official Claude Dev X Account\n\nQT @ClaudeDevs: For the developers building with Claude, a direct line from the team.

Follow for changelogs, API releases, community updates, and deep dives. https://t.co/SaH9KlMJ0z

See 3 related tweets

@chhddavid: RT @ClaudeDevs: For the developers building with Claude, a direct line from the team.

Follow for ch...

@_catwu: Excited to centralize our updates in one place!\n\nQT @ClaudeDevs: For the developers building with ...
@adocomplete: btw\n\nQT @ClaudeDevs: For the developers building with Claude, a direct line from the team.

Follow...

12. Forbes (Group Score: 171.4 | Individual: 32.5)

Cluster: 7 tweets | Engagement: 27 (Avg: 90) | Type: Tech

More than seven years since Forbes launched its first AI 50 list, the artificial intelligence industry has exploded, growing more expansive and increasingly too crowded for a single list to capture. As venture capital firms continue to pour money into AI, a new tier of startups has emerged: younger, earlier-stage companies building fast and raising faster as they try to rival their more established peers.

That’s why this year, for the first time, Forbes is introducing the AI 50 Brink List, spotlighting 20 of the most promising Seed and Series A-stage startups building in artificial intelligence.

Read more: https://t.co/MgjaxLFjBe #ForbesAI50 Photos: Nectar Social, Resolve AI, Periodic Labs, Ashley Maxwell, Giga, Jim Vetter, Studio B Portraits, Axiom

Sponsoring Partner @MayfieldFund

See 6 related tweets

@Forbes: AI isn’t slowing down—it’s accelerating.

Meet the AI 50 Brink List: 20 early-stage startups buildin...

@Forbes: 2026 #ForbesAI50

Artificial intelligence has become part of our lives, increasingly core to how we ...

@AppliedInt: Forbes AI 50 🤝 Applied Intuition\n\nQT @Forbes: 2026 #ForbesAI50

Artificial intelligence has become...

@brianzhan1: Traditional SaaS investors scoffed at the neolabs for their high valuations, and called it a bubble....
@glean: We’ve been named to the @Forbes AI 50 list for the fourth year in a row. 🎉

AI at work isn’t just ab...

13. Shipper_now (Group Score: 163.5 | Individual: 32.6)

Cluster: 6 tweets | Engagement: 26 (Avg: 24) | Type: Tech

Introducing Shipper

Claude Code Opus 4.6 can now build full-stack apps for you.

1️⃣ send your prompt in @shipper_now 2️⃣ opus designs, codes, launches, monetizes, translates, does emails 3️⃣ you go back to sleep and make $$$

Done. Claude is now your co-founder 🤝 https://t.co/uRnqWOFnUM

See 5 related tweets

@chhddavid: terrifying.\n\nQT @shipper_now: Introducing Shipper

Claude Code Opus 4.6 can now build full-stack a...

@chhddavid: apparently, Claude Code Opus 4.7 can now
build a full-stack app
launch it for me
do email mar...
@chddaniel: so you're telling me Claude Code Opus 4.7 can now...
build an app
launch it for me
do email m...
@chddaniel: life when claude code opus 4.7 is live https://t.co/oWHOzaKTZA\n\nQT @shipper_now: Claude Opus 4.7 w...
@chddaniel: RT @shipper_now: Claude Opus 4.7 was just released...

And now it is the default model in Shipper, t...

14. BrianRoemmele (Group Score: 157.3 | Individual: 34.7)

Cluster: 6 tweets | Engagement: 103 (Avg: 432) | Type: Tech

“Brian why are you so crazy about open source AI, you know companies have to make money right?”

Most of what you are reading this on is US OPEN SOURCE SOFTWARE with a proprietary layer.

Fact.

We don’t heed this we lose.

Fact.\n\nQT @BrianRoemmele: For those in the back row who think it is startup world in a 2012 replay with AI companies, not heeding the words of the very person you base your industry is foolish and will cripple the US.

LISTEN CLOSELY and if you work for and AI company make your boss listen. https://t.co/b0nRYOjqkx

See 5 related tweets

@BrianRoemmele: For those in the back row who think it is startup world in a 2012 replay with AI companies, not heed...
@BrianRoemmele: RT @BrianRoemmele: For those in the back row who think it is startup world in a 2012 replay with AI ...
@SeijinJung: openai started as a nonprofit AI safety research lab. they became a $300 billion company.

now they'...

@BrianRoemmele: RT @BrianRoemmele: “Brian why are you so crazy about open source AI, you know companies have to make...
@BrianRoemmele: Take 9 minutes and listen to the most sober and important statements about AI and the US position in...

15. MiniMax_AI (Group Score: 148.7 | Individual: 29.8)

Cluster: 6 tweets | Engagement: 110 (Avg: 184) | Type: Tech

Better agents are built together @NousResearch ❤️ @MiniMax_AI\n\nQT @MiniMaxAgent: Meet the world's first cloud sandbox Hermes Agent: MaxHermes 🪬

Every task it completes → new skills unlocked. Use it more → it gets better at being yours.

Start your Hermes era now: https://t.co/pmGzrLTWfk https://t.co/33Do9IZyEA

See 5 related tweets

@ivanfioravanti: I totally agree that Herms Agent + MiniMax M2.7 is match in heaven! 🚀\n\nQT @MiniMax_AI: Capable age...
@SkylerMiao7: MaxHermes is here. Run a self-improving agent in one click — cloud-hosted, secure, and frictionless....
@yaboilyrical: this is huge for terminally online open source agent framework users 🔥\n\nQT @MiniMax_AI: Capable ag...
@testingcatalog: MiniMax added a new always-on managed agent, powered by Hermes, to its MiniMax Agent platform.

Now...

@SkylerMiao7: Hermes is just the start. More agents powered by MiniMax are on the way to MiniMax Agent.\n\nQT @Min...

16. NotionHQ (Group Score: 139.4 | Individual: 54.3)

Cluster: 4 tweets | Engagement: 920 (Avg: 103) | Type: Tech

Opus 4.7, Anthropic's most intelligent model, is now in Notion!

It’s a real step up from Opus 4.6 for multi-step workflows. It uses fewer tokens. About 3× fewer tool errors. And it can troubleshoot weird mid-workflow issues like a real teammate. https://t.co/wg7HDdMTaP

See 3 related tweets

@GHchangelog: Claude Opus 4.7 rolls out on GitHub Copilot, improving multi-step tasks, long-horizon reasoning, and...
@jasonlk: Better ask your boss for more tokens:

“First, Opus 4.7 uses an updated tokenizer that improves how ...

@rickasaurus: RT @AndrewCurran_: Quoting: 'Opus 4.7 is a direct upgrade to Opus 4.6, but two changes are worth pla...

17. garrytan (Group Score: 133.0 | Individual: 45.3)

Cluster: 6 tweets | Engagement: 1013 (Avg: 213) | Type: Tech

Gemini Voice really is a cut above. It’s incredible.

Coming to GBrain and your OpenClaw/Hermes shortly\n\nQT @OfficialLoganK: Introducing Gemini 3.1 Flash TTS 🗣️, our latest text to speech model with scene direction, speaker level specificity, audio tags, more natural + expressive voices, and support for 70 different languages.

Available via our new audio playground in AI Studio and in the Gemini API! https://t.co/5PpBdhQMNg

See 5 related tweets

@rseroter: RT @demishassabis: Our most expressive and steerable TTS model yet! Designed to give builders granul...
@JeffDean: RT @OfficialLoganK: Introducing Gemini 3.1 Flash TTS 🗣️, our latest text to speech model with scene ...
@WesRoth: Google DeepMind unveiled Gemini 3.1 Flash TTS, an advanced text-to-speech model engineered for granu...
@rohanpaul_ai: RT @rohanpaul_ai: Google just launched Gemini 3.1 Flash TTS, a text-to-speech model that takes scene...
@unwind_ai_: RT @Saboo_Shubham_: Gemini 3.1 Flash TTS is here.

Our most expressive and controllable text-to-spe...

18. aakashgupta (Group Score: 131.4 | Individual: 35.7)

Cluster: 4 tweets | Engagement: 27 (Avg: 130) | Type: Tech

AI PM roles now pay $500K to$ 1.1M in total comp. OpenAI's average stock comp alone is $1.5M per employee per year.

The interview gating those offers didn't exist 18 months ago.

It tests whether you can draw a system architecture diagram, name your agents with non-overlapping responsibilities, pick XGBoost over GPT-4 for a prediction task and explain why, and surface failure modes before the interviewer asks.

The evaluation rubric weights technical fluency at 30-40%, system architecture at 25-30%, product judgment at 20-25%, trade-off articulation at 10-15%.

Most PM candidates spend 70% of the round on user personas. That's the 20-25% slice. They optimize for the smallest category and get rejected.

Hundreds of guides exist on product sense interviews. Zero on AI system design for PMs.\n\nQT @aakashgupta: Product design questions are dead.

The PM I just coached didn't get asked one in his AI PM job search.

Every round? AI system design.

First guide online: https://t.co/qgjmLPKkgf https://t.co/PgwCzH01ta

See 3 related tweets

@aakashgupta: Aman just landed an AI PM role.

Here's what he didn't face once in the loop: product design. "Desig...

@aakashgupta: AI PM comp at the top has gone absurd. At companies like OpenAI, Google, and Meta, yearly stock gran...
@aakashgupta: Product design questions are dead.

The PM I just coached didn't get asked one in his AI PM job sear...

19. StefanFSchubert (Group Score: 129.9 | Individual: 32.2)

Cluster: 5 tweets | Engagement: 368 (Avg: 59) | Type: Tech

RT @AlecStapp: This is the key moment between Jensen and Dwarkesh on export controls:

Dwarkesh asks why it’s okay to sell NVIDIA chips to China given the national security implications of AI models like Mythos.
Jensen gives a misleading answer, arguing that it’s okay to sell American chips to China because China already produces 60% of the world’s chips.
But as Jensen definitely knows, compute is measured in flops, not number of chips.
Dwarkesh then pushes back, pointing out that on a flops basis, China has 10% of the compute the US has, and giving them more compute would change their cyber capabilities.

This exchange shows why it’s critically important for interviewers to have at least some technical knowledge, so they can push back against misleading talking points.

See 4 related tweets

@ClementDelangue: Completely agree with Jensen!

Not exporting AI out of fear is classic case where the cure would be ...

@chandrarsrikant: RT @dwarkesh_sp: The Jensen Huang episode.

0:00:00 – Is Nvidia’s biggest moat its grip on scarce su...

@niccruzpatane: Possibly one of the hardest clips ever by a CEO. Jensen does not mess around.

"You don’t have to m...

@secureainow: https://t.co/r53FG1112v\n\nQT @AlecStapp: Here’s the one chart that shows Jensen is lying about the ...

20. kimmonismus (Group Score: 123.3 | Individual: 40.0)

Cluster: 5 tweets | Engagement: 490 (Avg: 373) | Type: Tech

OpenAI just dropped a major Codex update, one hour after Anthropic's Opus 4.7.

Whats new: background computer use on macOS (Codex clicks and types on your Mac while you keep working), in-app browser, image generation via gpt-image-1.5, persistent memory, long-running automations, and 90+ plugins including Atlassian, CircleCI, and Microsoft Suite.

The real signal is in OpenAI's own framing: 3M weekly Codex users, and they claim nearly half of usage is already non-coding. Codex is being repositioned as a general work agent, not just a dev tool. „Workbench for real work, before, around, and after writing code."

This is a different bet than Anthropic's. Opus 4.7 pushes the frontier on model capability. Codex pushes the frontier on what an agent does all day. Claude Code stays focused on software. Codex wants your whole workflow.

Frontier capability to Anthropic this week. Product surface area and distribution to OpenAI.

Anyways, big release probably coming thursday.

Spud\n\nQT @OpenAI: Codex for (almost) everything.

See 4 related tweets

@aakashgupta: The framing here tells you everything. "Workflows that don't expose an API."

OpenAI spent three yea...

@testingcatalog: OpenAI released Codex Super App 👀

Codex now has Memory, Computer Use, and browser capabilities. Loa...

@OpenAIDevs: RT @derrickcchoi: Really excited about today's @OpenAI Codex release that pushes it from a coding as...
@minchoi: RT @minchoi: This is wild.

OpenAI is just turned Codex into a full AI operator.

Mac apps, images, ...