科技热门推文——2026年8月21日

今日科技要闻聚焦人工智能领域的进展，涵盖开源智能体工具、大语言模型、无代码应用开发、专业法律模型，以及印度推动建设自主可控的数据基础设施。据报道，Anthropic计划首次公开募股，Callosum完成1亿美元融资，凸显投资者对该领域的强烈兴趣；与此同时，博通一项潜在规模达1000亿美元的人工智能融资方案，则反映出该行业不断攀升的资本需求。其他方面，Stripe因其工程实践被誉为行业标杆，而中国人形机器人产业则因高昂成本和商业可行性问题受到质疑。

1. hwchase17 (Group Score: 288.6 | Individual: 54.0)

Cluster: 10 tweets | Engagement: 436 (Avg: 42) | Type: Tech

RT @ataiiam: 🎉 Introducing 𝙾𝚙𝚎𝚗 𝙱𝚘𝚝

An open source Grok Bot that works with ANY agent harness, designed for real companies.

It includes:

AI Coworkers
Generative UI
Computer use (remote/local)
Agent-human handoffs
Full data recording, owned by you

Repo → https://t.co/ssje0KRts5

We're using this internally at @CopilotKit and it's changing the way we work forever.

More info below 👇

See 9 related tweets

@aiedge_: The best FREE Grok Bot directory on the entire internet.

Hundreds of Grok Bots that are ready to do...

@hwchase17: Open bot!\n\nQT @ataiiam: 🎉 Introducing 𝙾𝚙𝚎𝚗 𝙱𝚘𝚝

An open source Grok Bot that works with ANY agent ...

@RoundtableSpace: GROK BOT IS THE BEST AI AGENT RIGHT NOW.

IT GIVES YOU AN ARMY OF AGENTS THAT CAN WORK FOR YOU 24/7....

@sairahul1: OpenAI CEO Sam Altman:

"In the next 6 months, we're close to a world where a descendant of ChatGPT ...

@AnatoliKopadze: you can now try Grok Bot for free and i have no idea why this isn't everywhere

the only bottleneck ...

2. OfficialINDIAai (Group Score: 279.2 | Individual: 29.8)

Cluster: 14 tweets | Engagement: 1 (Avg: 4) | Type: Tech

WORKSHOP ON AIKOSH: Unlocking India’s AI Data

Data is central to building sovereign AI capabilities at scale.

The workshop examines how India can move from fragmented data resources towards a more interoperable ecosystem that enables innovation while keeping governance and trust at its core.

#AIKosh #IndiaAI #SovereignAI #AIForIndia

@AshwiniVaishnaw @JitinPrasada @SecretaryMEITY @sudeep_tweets @_DigitalIndia @GoI_MeitY @mygovindia

See 13 related tweets

@OfficialINDIAai: WORKSHOP ON AIKOSH: Unlocking India’s AI Data

The workshop brings together a diverse set of stakeho...

@OfficialINDIAai: WORKSHOP ON AIKOSH: Unlocking India’s AI Data

Trust is fundamental to unlocking the potential of da...

@OfficialINDIAai: WORKSHOP ON AIKOSH: Unlocking India’s AI Data

Building AI for India requires thinking deeply about ...

@OfficialINDIAai: Unlocking AI-ready data begins with enabling responsible access to the data India already holds.

Se...

@OfficialINDIAai: WORKSHOP ON AIKOSH: Unlocking India’s AI Data

Strengthening India’s AI & Digital Public Infrastruct...

3. Parul_Gautam7 (Group Score: 240.5 | Individual: 40.3)

Cluster: 8 tweets | Engagement: 419 (Avg: 302) | Type: Tech

I’ve seen a lot of AI benchmarks lately, but this angle genuinely caught my attention.

We’re used to asking, “Did the AI get the right answer?”

But for scientific discovery, sometimes there is no known answer yet.

That makes the process just as important as the result.

TRACES looks at things like tool use, mistakes, evidence, alternatives, and limitations — basically, whether the result was actually earned.

That feels like a much more practical way to think about trustworthy AI.\n\nQT @Apodex_AI: Most AI benchmarks test retrieval — can a model find the known answer? However, the hardest problems in science require discovery, can a system earn an answer nobody has yet?

Meet TRACES 🧭 — the world's first benchmark for measuring discoverative AI: AI that can work through evidence, test hypotheses, and reach verifiable conclusions on problems without answer keys. Proposed by our founder @tianqiao_chen, who defined its six capabilities.

Three things published today: a definition of "discoverative intelligence", a rubric to tell sound investigation from lucky guesses, and a open call for both solvers and problems *Website: https://t.co/iJ5rDV6qz9

See 7 related tweets

@heyshrutimishra: 90% of AI progress is an illusion built on benchmarks that reward memorization over thinking.

TRACE...

@AdarshChetan: AI can answer questions.

But can it make discoveries when the answer doesn’t exist yet?

That’s wha...

@Parul_Gautam7: RT @Apodex_AI: Most AI benchmarks test retrieval — can a model find the known answer? However, the h...
@heyshrutimishra: AI can write code in seconds, pass the bar exam, solve PhD problems but ask it something where the a...
@sheriyuo: TRACES: Each task declares which of six process abilities it can even exercise.

Latent ability cove...

4. robinebers (Group Score: 238.4 | Individual: 32.4)

Cluster: 12 tweets | Engagement: 3 (Avg: 761) | Type: Tech

stripe has been the continuous sleeper

i've got so much engineering and product experience, and stripe is the BEST example of how well things can go if you just have great teams caring

for months i've told everyone to ignore polar and all the others and literally just go straight to the source

stripe is the way, and openrouter is an S-tier addition

gg to both teams\n\nQT @OpenRouter: OpenRouter is joining Stripe.

We started OpenRouter with a simple mission: intelligence should be multi-model. Today, we are the largest AI marketplace & gateway, processing 10T+ tokens daily on 400+ models.

Joining @Stripe gives us the opportunity to accelerate that mission. https://t.co/21T5zZWfDM

See 11 related tweets

@AnandButani: Stripe just agreed to buy OpenRouter.

OpenRouter sits between your product and the model providers...

@tibo_maker: biggest takeaway:

Stripe billing is growing 70%+ Y/Y

Stripe is the internet economy and it's boo...

@Techmeme: Stripe's OpenRouter deal, its largest-ever acquisition, is a bet on a future where users turn to a m...
@novita_labs: A major new chapter for the multi-model AI ecosystem.

Proud to support OpenRouter as an inference p...

@Cointelegraph: 🔥 BIG: Stripe has agreed to acquire AI startup OpenRouter for over $7 billion, giving the payments g...

5. harvey (Group Score: 219.1 | Individual: 58.0)

Cluster: 6 tweets | Engagement: 1169 (Avg: 182) | Type: Tech

Introducing Tenet, our first model post-trained for legal.

Tenet is a Kimi K3 base that we post-trained with @FireworksAI_HQ on a corpus of publicly available legal data, synthetic data, and human expert data simulating long-horizon legal work.

Training increases Tenet's all-pass rate by 82% on LAB and 22% on LAB Contracts relative to the Kimi K3 base model. It achieves state-of-the-art performance on LAB Contracts and places second on LAB.

These gains generalize to other leading agentic benchmarks including @mercor's Apex Agents - Corporate Law, @crosbylegal's Redline Bench, and @scale_AI's Professional Reasoning Bench.

Tenet is also optimized for token efficiency, operating at less than a fourth the cost of leading foundation models.

We additionally post-trained three specialist models for Tenet to use as subagents:

M&A Diligence: post-trained with @baseten on our LAB Diligence environment in an RLM harness, this model is optimized for high-scale, long-horizon tasks.
Review Tables: trained with @appliedcompute on our Review Table environment, this model is state-of-the-art and cost-effective at high-volume document review and structured data extraction.
Firm Knowledge: trained with @EngramLab on our synthetic law firm environment, this model is optimized to learn and search over a firm's knowledge via memory and structured notes.

More details on model training, environment design, benchmarking, results, and more in the article by @gabepereyra below.

What's next for Harvey’s research?

Scaling LAB to more jurisdictions, practice areas and workflows
Scaling compute to bring new generalist models and capabilities to Harvey

More to come soon.\n\nQT @gabepereyra: https://t.co/1c2Dn0vuoX

See 5 related tweets

@gradypb: RT @harvey: Introducing Tenet, our first model post-trained for legal.

Tenet is a Kimi K3 base that...

@omarsar0: This is just the beginning of what's coming. Not paid to say this, but I think Harvey provides a gli...
@dejavucoder: waow harvey and neolab multi agent team release\n\nQT @harvey: Introducing Tenet, our first model po...
@harvey: RT @gabepereyra: Update on @harvey’s model training effort.

We post-trained a model we are calling ...

@harvey: RT @nikogrupen: An update on @harvey's post-training effort.

We’re introducing Harvey Tenet, which ...

6. NeuraFlowAix (Group Score: 208.8 | Individual: 26.1)

Cluster: 11 tweets | Engagement: 4 (Avg: 51) | Type: Tech

if you’re closing AI deals without opening the codebase, read this.\n\nQT @mardehaym: If you're reading this, I trust the algorithm targeted you because you evaluate software companies for a living. Experienced deal teams love saying they run thorough diligence.

I intend to show you why, for AI deals, they don't. https://t.co/q9nDwaXwRH

See 10 related tweets

@EmmaUsesAi: very important read before your next AI acquisition:\n\nQT @mardehaym: If you're reading this, I tru...
@alex_prompter: extremely important read if you evaluate AI companies:\n\nQT @mardehaym: If you're reading this, I t...
@NovaIAHQ: Very important if you evaluate software companies for a living:\n\nQT @mardehaym: If you're reading ...
@TheoBuildsAI: this is the most important article on AI M&A I’ve read this year.\n\nQT @mardehaym: If you're re...
@SynapseOpsAI: must read for every PE deal team running AI acquisitions.\n\nQT @mardehaym: If you're reading this, ...

7. exec_sum (Group Score: 193.4 | Individual: 33.7)

Cluster: 8 tweets | Engagement: 89 (Avg: 50) | Type: Tech

BREAKING: Anthropic is preparing to file for an IPO as soon as the end of August

The AI company reportedly expects the offering to match or exceed SpaceX's record-setting debut, which would make it one of the largest IPOs ever

See 7 related tweets

@StockMKTNewz: ANTHROPIC TO PREPARE IPO FILING AS SOON AS END OF AUGUST\n\nQT @IPONewsroom_: ANTHROPIC TO PREPARE I...
@wallstengine: SpaceX’s $SPCX IPO raised$ 75B, reaching $86.2B with the overallotment, the largest debut on record....
@Mayhem4Markets: Anthropic is expecting to file for its IPO as soon as the end of August.

It is expected to be an IP...

@kimmonismus: Anthropic is reportedly preparing to publicly file for its IPO as soon as the end of August.

The co...

@financialjuice: Anthropic is preparing for a potential IPO filing as soon as the end of August Anthropic CFO Krishna...

8. matthewclifford (Group Score: 182.6 | Individual: 37.4)

Cluster: 7 tweets | Engagement: 132 (Avg: 87) | Type: Tech

Callosum is one of the most ambitious companies I’ve worked with at @join_ef - incredible to see them raise $100m just a year after demo day. Testament to the scale of their opportunity and the speed of their execution - amazing work @achterbrain and @DanAkarca!\n\nQT @CallosumAI: Today we announce our$ 100M seed round to redefine how humanity computes in AI's next chapter.

The future of compute and AI is heterogeneous. Callosum is building it: https://t.co/t5KnOE0Sos

See 6 related tweets

@UKSovereignAI: A massive congratulations to @CallosumAI on raising a $100m seed round.

Callosum was our first inv...

@KanishkaNarayan: Britain is building the future of chips, the foundation for modern AI, an ambition shared by the @Ca...
@UKSovereignAI: RT @SebJohnsonUK: FULL INTERVIEW: I sat down with @DanAkarca to announce the 100m Seed.

Dan is a ge...

@UKSovereignAI: RT @CallosumAI: Today we announce our $100M seed round to redefine how humanity computes in AI's nex...
@Jameswise: RT @UKSovereignAI: A massive congratulations to @CallosumAI on raising a $100m seed round.

Callosu...

9. KyeGomezB (Group Score: 179.6 | Individual: 49.4)

Cluster: 5 tweets | Engagement: 602 (Avg: 90) | Type: Tech

RT @chamath: Texas, Pennsylvania and now Ohio.

This is a powder keg and has the potential to unwind 200-300 basis points of annual GDP if it metastasizes.

How did it get this bad?

The collective leadership of frontier AI has failed miserably in doing the basics:

Paint a positive picture of what is possible with AI. Instead of spilling the ink on the glass half full, stories abound of pearl clutching, catastrophizing or arcane technical jargon that, together, breed mistrust, fear and skepticism.
Visibly and disproportionately enrich the local communities AI data centers operate in. Tell these stories with a megaphone and have local leaders lead with the financial proof so it’s irrefutable.
Stop infighting and stop trying to influence the government or trick them because of the asymmetric knowledge the industry has relative to outsiders and observers. This further makes it seem like a shady money grab to many bystanders.

Data centers have unfortunately become THE symbolic representation of the asymmetric upside for a very narrow tech elite and a class that are untrustworthy.

Inasmuch, the safer bet may be that the pushback and resistance grows…

See 4 related tweets

@chamath: Texas, Pennsylvania and now Ohio.

This is a powder keg and has the potential to unwind 200-300 bas...

@Dan_Jeffries1: The backlash against data centers is directly the fault of AI leadership.

Too many folks have been...

@QuesoTrades: $CRWV uh oh, I JUST made a post about this yesterday. They have plans for Ohio data center in Clevel...
@MikeIsaac: smart and also trends with what i feel is a historically common belief among tech companies — "peopl...

10. wallstengine (Group Score: 178.9 | Individual: 31.4)

Cluster: 9 tweets | Engagement: 186 (Avg: 203) | Type: Tech

More details on $AVGO's reported AI financing:

The full package could reach nearly $100 billion, including$ 60B- $70B of senior secured debt plus roughly$ 30B of junior debt.

Apollo and Blackstone are in talks to participate, with the debt potentially issued through a special-purpose vehicle.

The financing would help Anthropic and other AI companies secure Broadcom chips and infrastructure, and could be rolled out in stages rather than all at once.\n\nQT @wallstengine: Broadcom $AVGO is reportedly seeking more than$ 60 billion in financing tied to its latest AI deal. The structure could include a $30 billion junior debt tranche, while Broadcom is also discussing guaranteeing part of the senior-secured tranche. https://t.co/WOuYnW5BdJ

See 8 related tweets

@ShanuMathew93: First NVIDIA, now AVGO getting in on the action

We’re going to figure out whether Accelerators beco...

@StockSavvyShay: $AVGO is reportedly seeking more than$ 60B in financing tied to a single AI deal.

The package could...

@wallstengine: Broadcom $AVGO is reportedly seeking more than$ 60 billion in financing tied to its latest AI deal. ...
@negligible_cap: *BROADCOM SEEKING MORE THAN $60 BILLION IN LATEST AI DEBT DEAL

That’s a whole lotta supply. At leas...

@edzitron: why does broadcom need this much debt? it's almost as if some of its customers can't afford its chip...

11. akiyoshisan (Group Score: 177.8 | Individual: 60.6)

Cluster: 5 tweets | Engagement: 1588 (Avg: 133) | Type: Tech

RT @ornith_: Aloha! 🌺Introducing Ornith-1.5, a family of open-source LLMs spanning 9B Dense, 35B MoE, and 397B MoE, trained with self-improving strategies.

It achieves state-of-the-art performance among open-source models of comparable size and delivers performance comparable to Claude Opus 4.8 across reasoning, agentic, and coding tasks: ✅Terminal-Bench 2.1 (86.1) ✅SWE-Bench (86 on verified, 65.1 on pro, 79.6 on Multilingual) ✅DeepSWE (56) ✅HLE (44.6) ✅ClawEval (81.4) ✅Tool Decathlon (71.2)

Ornith-1.5 takes a major step toward training foundation models through end-to-end self-improvement, extending the self-scaffolding strategies introduced in Ornith-1.0 into a more complete self-improvement loop: the model proposes new tasks, generates task-specific scaffolds, and produces solution rollouts for reinforcement learning, continuously creating new learning experiences from which it can improve.

All models, along with their quantized versions (FP8, GGUF, MLX, and NVFP4), have been released under the MIT License, enabling unrestricted commercial and research use. 📘Tech Blog: https://t.co/OZ63scRWLB 🤗Huggingface: https://t.co/mGJLwhrQOM

See 4 related tweets

@vllm_project: Ornith-1.5 is out and it serves in vLLM today. Congrats @ornith_ on a self-improving, MIT-licensed f...
@VaibhavSisinty: An open-source model is now matching Claude Opus 4.8.

Ornith-1.5 just dropped. Built by DeepReinfor...

@akiyoshisan: AIに「何をやらせるか」を人間が全部設計するところから、AI自身が新しい課題や学び方を作って、改善を回していく方向へ。

生成AIを毎日触っていると、モデル性能以上に「AIがどう成長していく仕組みを持...

@rohanpaul_ai: RT @rohanpaul_ai: Another brilliant open-source model release.

Ornith-1.5, a family of open-source ...

12. willccbb (Group Score: 165.7 | Individual: 37.1)

Cluster: 6 tweets | Engagement: 1173 (Avg: 217) | Type: Tech

honestly really weird that openai is perennially at 999M users and has revenue growth that looks more like databricks than anthropic despite being far better at product and marketing and developer advocacy and training useful cost-effective models. it really is a vibes economy

See 5 related tweets

@irl_danB: the curse of ChatGPT

OpenAI got a glimmer, a taste of the once per generation consumer company

so ...

@databricks: What does it take to run data behind products used by one billion people every week?

For @OpenAI, D...

@WesRoth: OpenAI reportedly just hit 20 MILLION active agent users.

Revenue run rate: +35% QTD Enterprise rev...

@cryptopunk7213: so openai’s agentic users are growing 7x faster than its revenue… huge volume that will be monetised...
@Yuchenj_UW: “OpenAI has revenue growth that looks more like Databricks than Anthropic.”

Databricks has database...

13. kimmonismus (Group Score: 161.0 | Individual: 27.7)

Cluster: 8 tweets | Engagement: 599 (Avg: 775) | Type: Tech

Damn : GPT-Astra still seems to be "weeks" away from release, and Anthropic presumably won't release Fable 5.1 before then eitherFeven though they could certainly use the better PR.\n\nQT @synthwavedd: 🚨 SCOOP: OpenAI yesterday told employees it intends to release Astra "in a couple [of] weeks" alongside demos of the model performing real-world tasks

An updated checkpoint, mostly an improvement to model alignment & reward hacking behaviours, is now being dogfooded (used internally by OpenAI employees via tools like Codex). Anthropic, meanwhile, are sitting on Fable 5.1 until Astra is launched. I think that's the wrong decision, but we'll see

See 7 related tweets

@mark_k: Astra is still a few weeks away!

🤷‍♂️\n\nQT @synthwavedd: 🚨 SCOOP: OpenAI yesterday told employees...

@haider1: openai looks like the slowest frontier lab over the last month

Fable 5.1 is already ready and proba...

@firstadopter: Oh.

"Q3 to date shows that OpenAI has surpassed Anthropic in QoQ enterprise growth: 82 vs. 76. Why...

@AndrewCurran_: RT @synthwavedd: 🚨 SCOOP: OpenAI yesterday told employees it intends to release Astra "in a couple [...
@MTSlive: "Anthropic hasn't released a new frontier model in months."

"If they were racing ahead, Fable 5.1 w...

14. Baconbrix (Group Score: 148.2 | Individual: 48.3)

Cluster: 5 tweets | Engagement: 1174 (Avg: 275) | Type: Tech

We've expanded Grok Build mode to all SuperGrok and X Premium users!

Create, remix, and prototype ideas from anywhere. No coding required 🚀\n\nQT @grok: Building apps on https://t.co/ISbWbJTky7, iOS, and Android is now available on every SuperGrok and X Premium plan

One prompt turns an idea into a published product with its own domain. https://t.co/5TxM2pP3V3

See 4 related tweets

@Baconbrix: New in the Grok app

Build and publish private apps to share exclusively with members of your team. ...

@grok: Building apps on https://t.co/ISbWbJTky7, iOS, and Android is now available on every SuperGrok and X...
@XFreeze: Grok Build is now available to all SuperGrok and X Premium users across iOS, Android and web

The ab...

@testingcatalog: Grok Build is rolling out to more SuperGrok users on web and mobile.

Testing time 👀 https://t.co/L...

15. MarioNawfal (Group Score: 146.8 | Individual: 32.0)

Cluster: 7 tweets | Engagement: 80 (Avg: 328) | Type: Tech

🇨🇳 China’s humanoid robot makers are packing Beijing’s World Robot Conference with flashy demos, but the real test is whether any of them can actually make money.

Units still cost up to $70k.

Brokerages figure a 2-year payback against human workers.

UBTECH’s humanoid revenue exploded more than 2,200% last year and the company still lost over $100M.

Analysts warn half or more of this year’s robots may end up gathering training data instead of doing paid work.

The hardware looks impressive, but still... the business model has a long way to go before it stops burning cash.

Writer: Lucas https://t.co/HbZFpCkmpb

See 6 related tweets

@MilkRoadAI: RT @MelvinInvests: China’s humanoid robot market is projected to grow 15x in five years and the whol...
@zijing_wu: Great one from Will. The Chinese version of circular financing:

Humanoid robot makers are generatin...

@Techmeme: China's humanoid robot makers derive much of their revenue from government-backed training centers t...
@business: Humanoid robots will reach “mass market” popularity within a decade, after manufacturers solve probl...
@CNBCTV18News: Unitree Robotics, the Chinese firm behind the robot dog Galgotias showcased at the India AI Summit m...

16. firstadopter (Group Score: 146.1 | Individual: 28.7)

Cluster: 7 tweets | Engagement: 79 (Avg: 125) | Type: Tech

True\n\nQT @jjohana: OpenAI will take the lead very soon in the Enterprise market. The joint release of GPT-5.6 Sol and Work was a game changer.

See 6 related tweets

@GavinSBaker: Yes. And the sum has accelerated slightly QTD with a sharper acceleration in August vs. July.

And n...

@scaling01: they call me leading indicator https://t.co/vVrvnwMB8m\n\nQT @arakharazian: May startle the markets ...
@TheRealAdamG: RT @jjohana: OpenAI will take the lead very soon in the Enterprise market. The joint release of GPT-...
@edzitron: I love this shit. Every day a new grenade\n\nQT @arakharazian: May startle the markets with this but...
@TheRealAdamG: RT @arakharazian: May startle the markets with this but: OpenAI is growing faster than Anthropic, ac...

17. robinebers (Group Score: 142.5 | Individual: 27.7)

Cluster: 6 tweets | Engagement: 35 (Avg: 761) | Type: Tech

best thing claude shipped in a year\n\nQT @ClaudeDevs: You can now set Claude Code's output style to Concise.

Claude leads with the result, keeps responses short, and still gives full detail when you ask.

Turn it on in /config → Output style, or set "outputStyle": "Concise" in settings.json. https://t.co/XYg7bHeVT2

See 5 related tweets

@adocomplete: Excited to try this tonight on a few projects.\n\nQT @ClaudeDevs: You can now set Claude Code's outp...
@ClaudeDevs: You can now set Claude Code's output style to Concise.

Claude leads with the result, keeps response...

@theo: Heads up: the /config command only edits the claude config for THE CURRENT PROJECT, NOT your global ...
@zzxwill: 嫌 Claude Code 废话篓子的朋友，可以配置一下。我自己反正需要\n\nQT @ClaudeDevs: You can now set Claude Code's output style ...
@cgtwts: WE ARE SO BACK 😭 https://t.co/elvc5wBCCk\n\nQT @ClaudeDevs: You can now set Claude Code's output sty...

18. SlackHQ (Group Score: 141.7 | Individual: 30.2)

Cluster: 7 tweets | Engagement: 516 (Avg: 101) | Type: Tech

RT @Benioff: Don’t code alone. Slack Code is live. Humans and agents. Same channel. Same work. Launching today with agents from @AnthropicAI, @github, @Cognition, and @vercel. This is real multiplayer coding. See it at @Dreamforce #DF26 https://t.co/scVeLgIgZR

See 6 related tweets

@Salesforce: RT @SlackHQ: Now building is a team sport.

Introducing Slack Code: a new way for your team and AI ...

@kimmonismus: It's interesting how Slack is developing. Most recently, there was the announcement of the deep inte...
@dani_avila7: Who needs an IDE when you have Slack?\n\nQT @Benioff: Don’t code alone. Slack Code is live. Humans a...
@SeijinJung: this is huge great work @SlackHQ team

my bet is Slack will continue to be the main interaction surf...

@Techmeme: Slack launches Slack Code, adding dedicated, project-specific channels that let teams collaborate wi...

19. teslaownersSV (Group Score: 136.0 | Individual: 46.8)

Cluster: 5 tweets | Engagement: 1263 (Avg: 215) | Type: Tech

BREAKING: Grok 4.6 just tied for #1 on the Artificial Analysis Agentic Index.⚡🤖

• Grok 4.6 (High) — 59 • Claude Opus 5 (Max) — 59

Grok 4.6 is now at the top of the benchmark, ahead of other leading models.

The Agentic Index focuses on capabilities that matter for real-world AI agents — tool use, planning, autonomy, and complex problem-solving.

This is especially significant as Grok expands into Grok Build and Grok Bot, where AI needs to do more than answer questions — it needs to use tools and complete tasks.

Source: Artificial Analysis

See 4 related tweets

@XFreeze: Grok 4.6 just tied for #1 on the Artificial Analysis Agentic Index

• Grok 4.6 (high) — 59 • Claude ...

@elonmusk: Grok 4.6\n\nQT @XFreeze: Grok 4.6 just tied for #1 on the Artificial Analysis Agentic Index

• Grok ...

@benitoz: I am having fun with Grok Bot it’s snappy, humorous, and gets stuff done now faster than Opus.

Thi...

@amuse: AI: For everyone that worried that Grok Bot would put you out of work realize that agentic AI will k...

20. alex_prompter (Group Score: 135.2 | Individual: 38.6)

Cluster: 4 tweets | Engagement: 91 (Avg: 47) | Type: Tech

very important read if you're a PE operating partner:\n\nQT @mardehaym: A PE firm asked us to rebuild an enterprise kitchen management platform for a company that powers chains like KFC.

We built it with one developer in six months.

Their Product Owner had vibe-coded the first version of the platform in Replit without any engineering background, and it worked well enough to win a pilot with a QSR chain running hundreds of locations.

Then the QSR chain needed it production-grade, certified, and deployed across hundreds of restaurants at enterprise scale. A vibe-coded Replit prototype wasn't going to get there.

We put one AI Velocity Pod on it: one developer and one PM.

Week one, we mapped the entire codebase before writing a line.

The PO had built working features, but the architecture underneath couldn't support what their contract demanded: multi-location deployment, role-based access across restaurant brands, and real-time demand forecasting.

We rebuilt the foundation and kept every working feature intact.

The PO kept building in Replit alongside our developer, which meant two parallel code streams feeding into one quality gate. Every line, whether agent-generated or from the PO's Replit sessions, got reviewed against production standards before it shipped.

In six months, one developer delivered:

Real-time demand forecasting and inventory management across hundreds of restaurant locations
Automated food safety labeling and compliance workflows
Cross-restaurant analytics and operational task management
Full multi-location deployment with role-based access, ready for enterprise QSR rollout

AI compute: $200 per developer per month.

A company that's been building kitchen technology for 20 years had one of our developers build their next platform.

The PE firm behind them has now sent us four portfolio companies, each one referred by the last CTO.

That's what the Velocity Framework was built for.

See 3 related tweets

@alex_prompter: the best engineering team on this app:\n\nQT @mardehaym: A PE firm asked us to rebuild an enterprise...
@alex_verem: very important read if you run portfolio companies:\n\nQT @mardehaym: A PE firm asked us to rebuild ...
@alex_prompter: RT @mardehaym: A PE firm asked us to rebuild an enterprise kitchen management platform for a company...