- Published on
热门科技推文 - 2026-06-23
- Authors
- Name
- geeknotes
科技每日简报 2026-06-23
Today's top tech conversations are led by @VaibhavSisinty, whose post about 'This is genuinely wild. 🤯
A J...' garnered the highest engagement. Key themes trending across the top stories include model, safety, built, anthropic, models. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.
1. VaibhavSisinty (Group Score: 993.3 | Individual: 65.8)
Cluster: 29 tweets | Engagement: 3937 (Avg: 243) | Type: Tech
This is genuinely wild. 🤯
A Japanese AI lab just built a model that matches Claude Fable 5 performance.
It's called Sakana Fugu, and it was built in Tokyo by a co-author of the original Transformer paper.
The guy who helped invent the architecture every AI runs on just built something that sits on top of all of them.
But here's the twist. Fugu isn't one giant model. It's a coordinator. You send one API call, and it decides which frontier model handles which part of the problem, runs them together, and hands back a single answer better than any of them could produce alone.
SWE-Bench Pro: Fugu Ultra scores 73.7. Claude Opus 4.8 scores 69.2. GPT-5.5 scores 58.6.
And because the model pool is swappable, if any provider gets export-controlled overnight, Fugu just routes around it.\n\nQT @SakanaAILabs: Introducing Sakana Fugu: A full multi-agent orchestration system accessible via a single model API.
Our ‘Fugu Ultra’ model matches the performance of Fable and Mythos, delivering frontier capability without the risk of export controls.
Try it: https://t.co/hhO6qTawgb 🐡
See 28 related tweets
- @hardmaru: Human intelligence is fundamentally a collective intelligence. We solve complex problems by particip...
- @Lux_Capital: RT @SakanaAILabs: Introducing Sakana Fugu: A full multi-agent orchestration system accessible via a ...
- @BullTheoryio: 🚨JAPANESE AI STARTUP JUST MATCHED CLAUDE FABLE 5 AND MYTHOS PERFORMANCE.
Japanese AI lab just launc...
- @testingcatalog: BREAKING 🔥: Sakana AI announced the Sakana Fugu and Sakana Fugu Ultra systems, which perform on par ...
- @chrisalbon: "Our ‘Fugu Ultra’ model matches the performance of Fable and Mythos, delivering frontier capability ...
2. Kyrannio (Group Score: 488.7 | Individual: 36.4)
Cluster: 17 tweets | Engagement: 704 (Avg: 97) | Type: Tech
RT @cb_doge: BREAKING: SpaceX will be paid ~$2.32 billion per month from these 3 AI compute deals.
• Anthropic: $1.25B/month
• Google: $920M/month
• Reflection: $150M/month
That’s a nearly $28 billion annual run rate from AI compute alone. https://t.co/G3oZxkfTs5
See 16 related tweets
- @business: SpaceX signed a multibillion-dollar deal to provide computing power to AI startup Reflection AI, the...
- @IPONewsroom_: SPACEX 6.3B
Open-source AI startup Reflec...
- @WOLF_Financial: SpaceX 6.3B\n\nQT @IPONewsroom_...
- @TFTC21: SpaceX just signed a $6.3 billion compute deal with AI startup Reflection for capacity at Colossus 2...
- @XFreeze: SpaceXAI just signed another major compute deal
Reflection AI, an Nvidia-backed open-source AI star...
3. firstadopter (Group Score: 440.9 | Individual: 39.3)
Cluster: 16 tweets | Engagement: 129 (Avg: 64) | Type: Tech
MULTI-YEAR GROWTH SUPPLY AGREEMENT DRAM
"Building on the technical collaboration, Micron and Anthropic have entered into a memory and storage supply agreement spanning Micron's industry-leading data center portfolio. This positions Micron to support Anthropic’s multi-year growth trajectory as the frontier AI lab scales its compute strategy for the long term."\n\nQT @firstadopter: June 22, 2026 at 9:00 AM EDT Micron and Anthropic Announce Strategic Agreement to Scale Next-Generation AI Infrastructure Collaboration spans memory and storage AI architecture design, supply agreement, enterprise AI adoption and a strategic investment BOISE, Idaho, June 22, 2026 (GLOBE NEWSWIRE) -- Micron Technology, Inc. (Nasdaq: MU) today announced a strategic agreement with Anthropic that spans memory and storage AI architecture design, supply and demand, enterprise adoption of Claude across Micron and a strategic investment in Anthropic’s Series H funding round. The agreement directly links the demands of frontier AI models to how infrastructure is designed, supplied, and deployed at scale.
See 15 related tweets
- @zephyr_z9: Bank of Memory is open for business Hynix, Micron, & Sammy are going to acquire a lot of assets ...
- @StockMKTNewz: MICRON AND ANTHROPIC JUST ANNOUNCED A NEW AGREEMENT
Micron $MU and Anthropic just announced a new a...
- @danielnewmanUV: $MU leaning into Anthropic. The ‘commodity’ is now a ‘strategic’ in the AI boom. 💪🏻👀\n\nQT @MicronTe...
- @wallstengine: Micron $MU signed a strategic agreement with Anthropic covering AI memory/storage architecture desig...
- @dejavucoder: https://t.co/JfLptGLzeN\n\nQT @firstadopter: June 22, 2026 at 9:00 AM EDT Micron and Anthropic Annou...
4. chandrarsrikant (Group Score: 336.2 | Individual: 49.1)
Cluster: 15 tweets | Engagement: 3486 (Avg: 225) | Type: Tech
BREAKING: Moneycontrol scoop onMeta and CRED confirmed AND ITS EVEN BIGGER Kunal Shah to become new global CEO of WhatsApp, replacing Will Cathcart Meta to invest a billion dollars in CRED
See 14 related tweets
- @business: Meta is investing $900 million into Indian fintech startup Cred, with plans to appoint its founder, ...
- @wallstengine: 900M in Indian fintech startup Cred for a ~20% stake, valuing them at $4.5B post...
- @TechCrunch: WhatsApp gets new chief as Meta taps India’s CRED founder Kunal Shah, and invests $900M in startup h...
- @chandrarsrikant: 🚨🚨 Biggest startup story of the year that broke right here on Moneycontrol Meta to invest nearly $1...
- @chandrarsrikant: 🚨What Kunal Shah's appointment as WhatsApp global head means for Meta and India
Cred founder Kunal ...
5. lukas_m_ziegler (Group Score: 262.1 | Individual: 37.4)
Cluster: 12 tweets | Engagement: 131 (Avg: 209) | Type: Tech
JUST IN: NVIDIA just announced Halos for Robotics, the industry's first full-stack safety system for Physical AI.
And it's built on 18,600+ engineering years of autonomous vehicle safety development.
This is a big deal. Here's why.
Humanoid robots are about to operate alongside humans on real factory floors. In warehouses. In logistics operations. The single biggest barrier to that happening at scale is not capability, it's safety certification.
How do you prove a robot is safe enough to work next to a person?
Halos for Robotics spans every layer of the stack:
→ NVIDIA IGX Thor, industrial-grade AI compute with built-in safety capabilities → Halos OS, full software stack for robotics safety functions → Outside-In Safety Blueprint, external cameras and AI agents that dynamically control robot behaviour in real time → NVIDIA Halos AI Systems Inspection Lab, the world's first ANAB-accredited programme for functional and AI safety for physical AI
Agility Robotics is the first adopter, integrating Halos directly into Digit, its humanoid already deployed with Amazon, GXO, Schaeffler and Toyota.
The certification bodies involved read like a who's who of industrial safety, TÜV Rheinland, TÜV SÜD, UL Solutions, SGS and more. Over 40 companies in the ecosystem already.
First it did it for AI. Then for autonomous vehicles. Now for robotics.
Whoever owns the safety infrastructure layer owns the keys to humanoid deployment at scale.
@NVIDIARobotics
~~
♻️ Join the weekly robotics newsletter, and never miss any news → https://t.co/GoA3ZuwoPB
See 11 related tweets
- @NVIDIARobotics: Deploy safe robots without rebuilding safety from scratch. 🦾
NVIDIA Halos for Robotics extends the ...
- @StockMKTNewz: NVIDIA $NVDA JUST LAUNCHED THE FIRST FULL-STACK SAFETY SYSTEM FOR ROBOTICS AND PHYSICAL AI
It is ca...
- @DeItaone: $NVDA - NVIDIA PUSHES SAFETY LAYER FOR HUMANOID ROBOTS
Nvidia is developing software and chips to i...
- @nvidianewsroom: NVIDIA announces NVIDIA Halos for Robotics, the industry’s only full-stack, open robotics safety sys...
- @humanoidsdaily: RT @humanoidsdaily: For humanoid robots to move out of cages and onto real factory floors, safety ca...
6. MTSlive (Group Score: 241.3 | Individual: 32.7)
Cluster: 11 tweets | Engagement: 51 (Avg: 110) | Type: Tech
SITUATION EXPLAINED: Google just lost two of the most important AI researchers in the world in the same week.
• Noam Shazeer, co-author of "Attention Is All You Need" and co-lead of Gemini, left for OpenAI • John Jumper, Nobel laureate in Chemistry 2024 for AlphaFold, left for Anthropic • Google stock down 6% today • There are roughly 50 people in the world who truly understand how to train a frontier model end-to-end • These people are likely being paid Ronaldo/Messi money, 300M per year range
@theojaffee: "By the way, we don't actually believe in the downfall of Google. AI leads change spots so often, like once a month."
See 10 related tweets
- @Jessicalessin: It is interesting to me that major tech companies' stocks still rise and fall off AI talent coming o...
- @firstadopter: Things are getting serious https://t.co/Lz7P5p0nEQ\n\nQT @firstadopter: The smartest money in the wo...
- @firstadopter: The smartest money in the world piled into Google as an AI winner. I'm not being sarcastic. The smar...
- @TFTC21: Google is down 6.7% today to one-month lows.
On Polymarket, the company has just an 11% chance of ...
- @firstadopter: Just one month ago, the consensus among investors and financial media was that Google was leading in...
7. jxnlco (Group Score: 225.4 | Individual: 37.7)
Cluster: 9 tweets | Engagement: 526 (Avg: 134) | Type: Tech
RT @OpenAI: We’re expanding OpenAI Daybreak to help democratize patching vulnerable software at machine speed:
Codex Security plugin: find, validate, and fix vulnerabilities right inside Codex
The full version of GPT-5.5-Cyber model: a great model for trusted defenders
Cyber Partner Program: powering products built on top of our best cyber capabilities for leading security companies to secure the world's software
Patch the Planet: working with maintainers to secure critical open source projects
See 8 related tweets
- @AiBattle_: The new GPT-5.5 Cyber model achieves a higher CyberGym score than Mythos 5
Among open-source models...
- @TheRealAdamG: https://t.co/WHK1aL5791
"We’re expanding @OpenAI Daybreak to help democratize patching vulnerable ...
- @MTSlive: SITUATION UPDATE: OpenAI has launched the full version of GPT-5.5-Cyber, its most capable cybersecur...
- @TeksEdge: Seems like OpenAI is following Anthropic’s strategy of pre-releasing powerful AI models early (often...
- @SIGKITTEN: Patch the Planet https://t.co/zdRpQlByRL\n\nQT @OpenAI: We’re expanding OpenAI Daybreak to help demo...
8. rohanpaul_ai (Group Score: 223.1 | Individual: 33.2)
Cluster: 7 tweets | Engagement: 20 (Avg: 54) | Type: Tech
Delos just launched Workers.
Turns AI agents into always-available digital employees with their own email, phone, Slack, Teams, computer, memory, and company-specific training.
Delos says each worker is trained on one business, so the system learns the company’s routines, tools, vocabulary, permissions, and handoff patterns instead of acting like a generic assistant dropped into a random workplace.
The key point is the persistent identity plus long-term memory, because an AI employee only becomes useful if people can contact it like a coworker and trust it to remember past decisions, open loops, customer context, and internal procedures.\n\nQT @pierre_dlgr: Introducing Workers: unlimited AI employees that run your company...
And we've just made $1M ARR in a couple of days.
Most knowledge jobs shouldn't exist.
Not because the work isn't valuable.
Because the way we execute it is absurd.
Millions of people spend their days:
- Answering emails
- Following up
- Updating CRM
- Coordinating meetings
- Moving information between systems
There is no expertise here. That's workflow..
Workers are AI employees that can perform work across your organization like a human teammate.
The line between software and labor just disappeared.
See 6 related tweets
- @FellMentKE: We're moving from AI tools that assist with tasks to AI systems that can actually participate in wor...
- @kimmonismus: Most AI agents reset the moment a task ends: you prompt, it runs, the context is gone.
Delos Worker...
- @TheoBuildsAI: Imagine you're a VC.
A founder comes to see you in 2024. "I want to create AI employees capable of ...
- @dr_cintas: “Workers” just launched 🤯
AI employees that run tasks across your whole company, like a human teamm...
- @cgtwts: POV: your newest employee works 24/7, never misses a deadline, and costs less than Slack. https://t....
9. chandrarsrikant (Group Score: 221.7 | Individual: 39.7)
Cluster: 8 tweets | Engagement: 470 (Avg: 225) | Type: Tech
Contours of CRED and Meta deal, now confirmed and official..Meta says it won't receive access to Cred customer information
CRED will raise ₹8,550 crore (~US$900 million) in its Series H funding round led by Meta
Investment will be structured through a mix of primary and secondary share purchases.
Valuation
Post-money valuation: ₹43,239 crore (~US$4.5 billion).
Pre-money valuation: ₹38,819 crore (~US$4.03 billion).
Meta's Investment
Meta will join CRED's cap table as a minority investor.
Meta will not receive access to CRED customer information.
CRED Business Highlights
Serves 1.7 crore (17 million) members monthly.
Offers products across:
Payments
Lending
Insurance
Wealth management
Lifestyle services
Processes over 40% of credit card bill payments in India.
Has built the category over the past 8 years.
Lending business has grown to ₹24,000 crore (~US$2.5 billion+) managed AUM for leading financial institutions.
Use of Funds
Accelerate growth.
Strengthen organizational and institutional capabilities.
Expand leadership across existing and new categories.
Leadership Changes
Kunal Shah will step down from his operating role as CEO.
He will retain his personal shareholding in CRED.
Kunal will transition to Meta's global leadership team.
Miten Sampat will become Interim CEO, effective immediately.
The board is working on a long-term leadership structure in preparation for an eventual IPO.
Kunal Shah's Key Remarks
Founded CRED in 2018 to reward creditworthy consumers.
CRED has grown into:
A new financial services category.
A platform with millions of members.
A business generating approximately ₹3,200 crore (~US$325 million) in revenue.
A profitable company with a full stack of licenses and a strong brand.
Expressed confidence in the team’s ability to continue scaling the company.
See 7 related tweets
- @chandrarsrikant: Omg omg omg Meta Investment in CRED is now confirmed and Kunal Shah will be heading WhatsApp global...
- @Rajananandan: Congratulations @kunalb11 @miten and entire #Cred team! Excited to see what you do with @WhatsApp #K...
- @pitdesi: Very interesting- single person acquihire sorta
Meta invests 4.5B...
- @moneycontrolcom: #StartupsWithMC | Meta won't get access to Cred users' financial data despite $900 million investmen...
- @Rajananandan: RT @miten: In 2020 when I joined @kunalb11 at CRED, I was drawn to the incredible team and a contra...
10. Kyrannio (Group Score: 210.8 | Individual: 46.1)
Cluster: 7 tweets | Engagement: 848 (Avg: 97) | Type: Tech
RT @SakanaAILabs: Fugu stands shoulder-to-shoulder with leading models like Fable and Mythos across the industry's most rigorous engineering, scientific, and reasoning benchmarks.
Read the full blog: https://t.co/2ZJbdWqCUj
Beyond Bigger Models: Why are Orchestration Models the Next Frontier
Progress in AI has been driven largely by giant, monolithic models. But the most powerful systems of the future will be collaborative ecosystems.
Today, this orchestration is no longer just a technical optimization. It has become a geopolitical and operational imperative.
For an organization or a nation, relying on a single company's model for critical infrastructure, finance, or governance is a material vulnerability. This risk is no longer a hypothetical possibility, but a reality.
As we have seen with recent export controls imposed on models like Fable and Mythos, access can disappear overnight.
Collective intelligence is the practical hedge against this concentration of power. Because Fugu orchestrates an underlying pool of swappable agents, it simply routes around vendor restrictions.
By orchestrating the world’s models, we are delivering the resilient blueprint required for true AI sovereignty.
See 6 related tweets
- @KyeGomezB: RT @hardmaru: Human intelligence is fundamentally a collective intelligence. We solve complex proble...
- @ivanfioravanti: "Since we started Sakana AI, our core conviction has been that the most powerful AI systems will be ...
- @vkhosla: More than one way to get to to SOTA performance . Exciting.\n\nQT @SakanaAILabs: Fugu stands shoulde...
- @BrianRoemmele: Once again busting the fable of Fable.
This company from Japan is getting near to run circles aroun...
- @Scobleizer: Big claims to open the week on.\n\nQT @SakanaAILabs: Fugu stands shoulder-to-shoulder with leading m...
11. wallstengine (Group Score: 191.4 | Individual: 28.5)
Cluster: 11 tweets | Engagement: 103 (Avg: 109) | Type: Tech
75M in A24 as part of a multi-year AI research partnership between Google DeepMind and the film studio.
The deal marks Google’s first stake in a studio.
A24 and DeepMind plan to build AI tools for movie production and distribution, while the agreement does not give Google access to A24’s film and TV library.
A24 Labs is already developing an AI-generated storyboard tool.
See 10 related tweets
- @MTSlive: SITUATION DETECTED: Google is investing $75 million in A24 as part of an AI research partnership, pe...
- @MorePerfectUS: Google is investing $75 million into movie company A24 as part of an AI research partnership.
The t...
- @WSJ: Exclusive: Google is investing in independent movie studio A24 as part of a new AI research partners...
- @StockSavvyShay: 75M in A24 marking its first film studio stake through a multi-year AI researc...
- @WOLF_Financial: Google 75 Million into movie studio A24\n\nQT @PolymarketMoney: BREAKING: $GOOGL...
12. PandaTalk8 (Group Score: 153.5 | Individual: 54.8)
Cluster: 4 tweets | Engagement: 482 (Avg: 32) | Type: Tech
RT @alisawuffles: I'm joining OpenAI next week!🥹 The job search turned out to be really challenging but also super rewarding, so I wrote a small blog to share what I learned along the way and hopefully make the process a little less mysterious for the next person. https://t.co/6FigSBdenD
See 3 related tweets
- @Hesamation: this PhD student had 47 interviews and 4 offers before she was hired at OpenAI.
she practiced with ...
- @sheriyuo: This blog is really helpful for fresh graduates navigating the job search.
Thanks for sharing what...
- @cgtwts: This PhD student did 47 interviews and received 4 offers before landing at OpenAI.
Her prep notes o...
13. business (Group Score: 152.4 | Individual: 23.8)
Cluster: 9 tweets | Engagement: 169 (Avg: 81) | Type: Tech
Chevron signed a 20-year deal with Microsoft to provide natural-gas fired power for a proposed West Texas data center, which could be one of the biggest in the US https://t.co/ipQU4mXxeG
See 8 related tweets
- @mollytaft: reported on this project in May, as it's seeking millions in state tax breaks that are normally set ...
- @StockMKTNewz: Microsoft CVX have agreed to a 20 year deal to provide
Natural-gas fired power ...
- @WOLF_Financial: Microsoft CVX to supply energy to one of its proposed...
- @Polymarket: JUST IN: Chevron signs 20-year deal with Microsoft to supply natural gas power to its Texas data cen...
- @unusual_whales: JUST IN: Chevron, MSFT, to provide natural-gas fired po...
14. _simonsmith (Group Score: 135.2 | Individual: 28.0)
Cluster: 6 tweets | Engagement: 16 (Avg: 32) | Type: Tech
GLM-5.2 seems like a good model, but the fact it can achieve #3 on GDPval-AA while being a TEXT ONLY MODEL suggests GDPval-AA doesn't reflect a lot of knowledge work, which requires reading charts and figures, tweaking designs, and other visual activities.\n\nQT @ArtificialAnlys: GLM-5.2 leads open weights models and sits at #3 overall on GDPval-AA, a real-world agentic work benchmark
GLM-5.2 from @Zai_org scores 1524 Elo on GDPval-AA, which measures performance on real-world, economically valuable knowledge work through long-horizon, multi-turn tasks.
Key takeaways:
➤ #3 overall, behind only Claude Fable 5 (1783) and Claude Opus 4.8 (1615), and level with GPT-5.5 (xhigh, 1509)
➤ The leading open weights model by a wide margin: the next open model, MiniMax-M3, scores 1408
➤ Ahead of many proprietary models, including Google's Gemini 3.5 Flash (1357), Qwen 3.7 Max (1289), Muse Spark (1158)
➤ The tasks are agentic. GLM-5.2 averaged ~31 turns per task across 1,999 matches
➤ Consistent with the rest of its launch, GLM-5.2 also leads open weights on the Artificial Analysis Intelligence Index, ranks #3 on the Agentic Index, and #3 on AA-Briefcase
See 5 related tweets
- @aiedge_: GLM-5.2 is the most powerful open-sourced AI model ever.
This article teaches you:
- what it's goo...
- @ClementDelangue: RT @ArtificialAnlys: GLM-5.2 leads open weights models and sits at #3 overall on GDPval-AA, a real-w...
- @Zapier: RT @wadefoster: Open-source models are only 3 months behind frontier models.
GLM 5.2 dropped last w...
- @WesRoth: GLM-5.2 is now the highest-scoring open-source model on the DeepSWE v1.1 coding-agent benchmark.
At...
- @AravSrinivas: GLM is the kind of model that revives serious interest in open source AI. It passes the blind test r...
15. jenzhuscott (Group Score: 131.9 | Individual: 35.1)
Cluster: 4 tweets | Engagement: 376 (Avg: 170) | Type: Tech
RT @eliebakouch: to be clear, this is a closed source orchestrator on top of closed source models. if before you didn't control the models, now you don't even control which ones are used or how much. this is not "AI sovereignty"
i've also read the tech report to get an opinion on the technical stuff:
fugu (not the ultra version) is basically a classifier that selects which model at each turn is most likely to answer correctly (in other words a router). this leads to -10 points on SWE Bench pro compared to opus, gets some gains on other benchmarks but very slight. argument could be that it reduces cost, but no information about this so it's likely the opposite. they also have an autoresearch benchmark where they compare to frontier models "Model A, B and C" which is really crazy to not be transparent about what models you compare against. let's also say that this probably doesn't support adding new llm out of the box since you need to retrain the classifier
about fugu ultra, this is basically and advanced plan mode and orchestrator, this is a model that for a query outputs a plan with multiple "workflows". my understanding of workflows is that they say: "spawn model A subagents to achieve this, then use model B to judge it, then summarize this with model C" which is just a test time scaling compute strategy. i think this is an okish way to do it, but it's limited by the fact that they need to predict everything before the agents start working, which is why they limit this to 5 steps. imo you need to predict what to spawn at t+1 with the information you get at t, not with the info you get at t=0. there are also other issues such as fable 5 score on terminal bench being wrong and them being super vague and unclear about which model is in the LLM pool (they only mention closed source api one)
the biggest and most obvious issue is that they are introducing a "test time scaling" method with "best of N" over models, and they literally NEVER REPORT the number of output tokens or cost to achieve a benchmark/task
the good comparison here is not with opus, but it's opus with ultracode/workflows enable, not with kimi, but with kimi swarm ect.. very very confusing release
See 3 related tweets
- @jenzhuscott: A sober look into the @SakanaAILabs release.
It uses dynamic routing and is model-agnostic, which ...
- @diptanu: The market is hungry for a Mythos level model.\n\nQT @eliebakouch: to be clear, this is a closed sou...
- @didier_lopes: Definitely test-time compute scaling
E.g. for multi-turn agentic tasks they see Fugu Ultra routing ...
16. VKazulkin (Group Score: 118.1 | Individual: 55.2)
Cluster: 3 tweets | Engagement: 1238 (Avg: 109) | Type: Tech
RT @0xMovez: Andrew Ng:
"100% of my tasks are now done by AI agents - hype has exceeded my expectations. Loops is next step.
in 3-6 months, everyone will be using self-improving loops. No more prompting."
In a 30-minute talk, Andrew Ng explains how to build self-improving agentic systems from scratch.
Worth more than a $500 agentic course.
See 2 related tweets
- @AnatoliKopadze: Google Brain founder, Andrew Ng:
"100% of my tasks are done by ai agents, self-improving loops are ...
- @mikenevermiss: Andrew Ng:
“100% of my tasks are now done with ai agents. the progress has exceeded my expectations...
17. chandrarsrikant (Group Score: 111.3 | Individual: 46.4)
Cluster: 4 tweets | Engagement: 3960 (Avg: 225) | Type: Tech
Leave Cred aside for a moment A startup founder from India will now lead WhatsApp globally Unprecedented
See 3 related tweets
- @chandrarsrikant: 🚨Can Kunal Shah turn WhatsApp's scale into revenue?
At CRED, founder Kunal Shah built a business ar...
- @FirstSquawk: MUMBAI INDIA BASED CRED FOUNDER KUNAL SHAH APPOINTED GLOBAL CEO OF WHATSAPP IN LANDMARK META SHAKE-U...
- @BBCWorld: WhatsApp to be led by Indian start-up founder as Will Cathcart steps back https://t.co/zs4oh3NsUq...
18. peterwildeford (Group Score: 110.9 | Individual: 44.7)
Cluster: 4 tweets | Engagement: 215 (Avg: 60) | Type: Tech
It makes sense that a new Mythos would be available by now, after all it's been four months since the original Mythos finished training! Four months is a long time in AI and nothing about the recent actions by the Trump administration actually pauses AI development internally.
If current AI trends for 2024-2026 continue, this new Mythos likely scores >4h on the "METR 80%" and has a decent shot of >6h (current-Mythos was 3.1h and Claude 4.6 Opus was 1.2h).
The timing also makes sense. Anthropic-Amazon New Carlisle opened in Dec 2025 with 471k H100-equivalents of compute and Mythos came out three months later.
In March, Anthropic-Amazon New Carlisle was upgraded to host 687k H100-equivalents and June would be about three months after that as well. With twice as much compute as before, new-Mythos is probably quite capable!
Like I always said, Mythos was just the beginning and the AI trend is continuing. Whether we're ready for that is increasingly a guess rather than a certainty.\n\nQT @AndrewCurran_: A new, more capable version of Mythos has emerged from training. I don't know whether it will be called Mythos 5.1 or Mythos 6, or if Anthropic will keep it internal to accelerate further development - but it has arrived.
Stopping models like Fable 5 or Mythos 5 from being served to the public does nothing to slow down development. In fact, it probably speeds it up slightly by freeing up resources. There are also no rules preventing the labs from continuing to advance capabilities while any current model is under embargo - or from keeping progress quiet until they choose to release it. None of them can afford to pause or slow down. We need only look at how capable GLM-5.2 is as proof of this. To protect their business models, the frontier labs must continually train increasingly capable systems to stay ahead of open source, and each other. The current continues to rage beneath the ice, and we continue to race toward our destination.
See 3 related tweets
- @tengyanAI: if you keep the best models to yourself and away from the public + competitors (whether by choice or...
- @scaling01: Anthropic having a new Mythos version is not surprising
Original Mythos was likely trained between ...
- @AndrewCurran_: RT @peterwildeford: It makes sense that a new Mythos would be available by now, after all it's been ...
19. cgtwts (Group Score: 107.9 | Individual: 48.1)
Cluster: 3 tweets | Engagement: 1736 (Avg: 307) | Type: Tech
Sir… Sakana just launched their own Fable. https://t.co/TiaqyvDVKw\n\nQT @SakanaAILabs: Introducing Sakana Fugu: A full multi-agent orchestration system accessible via a single model API.
Our ‘Fugu Ultra’ model matches the performance of Fable and Mythos, delivering frontier capability without the risk of export controls.
Try it: https://t.co/hhO6qTawgb 🐡
See 2 related tweets
- @rileybrown: I tried this. Tried to make a landing page to test its design skills as my first prompt. It didn't e...
- @Hesamation: “delivering frontier capability without the risk of export controls” well EU has an import control p...
20. NousResearch (Group Score: 106.2 | Individual: 32.1)
Cluster: 4 tweets | Engagement: 2395 (Avg: 528) | Type: Tech
Hermes Agent has reached 200,000 GitHub stars
Thank you to our contributors, supporters, users, and agents! https://t.co/4Oca1XO7H7
See 3 related tweets
- @Teknium: We made it to 200K stars, let’s go!!\n\nQT @NousResearch: Hermes Agent has reached 200,000 GitHub st...
- @iamlukethedev: 200,000 GitHub stars.
That is not just a milestone.
That is a signal.
Hermes Agent went from “int...
- @tunguz: https://t.co/72NrJdAeMB\n\nQT @NousResearch: Hermes Agent has reached 200,000 GitHub stars
Thank yo...