Sebastian Werner @swernerx

@MengTo fable 5 is basically a frontend intern with taste now

I'm having a lot success using Fable xhigh as a planner/architect, using GPT 5.5 xhigh (subscription) as a coder, then Fable xhigh again as a judge. At API pricing, planning+judge costs are in the ~few dollar range compared to typical $50+ full round trips. I've seen some others using dumber/cheaper coders, but GPT 5.5 even at xhigh compared to Fable 5 is very cheap and very fast. And GPT 5.5 is just... really good. Still been less than 24hrs since the re-release so the longevity of this approach is unclear, but its been working really well.

Claude Fable 5 will be available again globally tomorrow. After a series of productive conversations with the US government, we're redeploying the model with a new set of classifiers to target and block more cybersecurity tasks. In the near term, some routine tasks like coding and debugging will fall back to Opus 4.8. We’ll continue to refine these classifiers over the coming weeks to reduce false positives and better distinguish genuine misuse from legitimate requests. We’ve also begun drafting a consensus framework—with Amazon, Microsoft, Google, and other Glasswing partners—for assessing the severity of AI jailbreaks and how AI developers should respond to them. We invite other industry partners and model providers to join us in this effort. Finally, we’re scaling up our collaboration with the US government on model testing and safeguards. This will include pre-release access to models and safeguards for evaluation, information sharing on jailbreaks and misuse, and dedicated resources for joint research. Thank you to our users for your patience, and to our partners across the government, industry, and the research community who worked alongside us to make Fable 5 available again. Read our full blog: anthropic.com/news/redeployi…

We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5. We'll begin restoring access tomorrow, and will share an update soon. We’re grateful to our users for their patience, and to everyone who worked with us on redeploying the models.

This is actually wild. Hermes just let you merge any two AI models into one virtual model. 🤯 It is called Mixture of Agents. Here is how it works. You pick any two models. GPT-5.5 and Claude Opus for example. One runs as the reference, one as the aggregator. Name the combo anything you want. It shows up as a single selectable model in your picker like any other. Every task, both models run in parallel. The reference analyzes and responds. The aggregator reads that, synthesizes everything, writes the final answer, and handles all tool calls. You see one clean output. The results on hard agentic tasks: → 8% higher than Opus 4.8 alone → 11% higher than GPT-5.5 alone Full Hermes features work untouched. Memory, tool use, skills, long sessions, cross-channel messaging. Nothing breaks. The combo just performs better than either model on its own. You can mix any providers too. OpenAI, Anthropic, OpenRouter, local models. Whatever you have access to.

Nous Research @NousResearch

2 weeks ago

The strongest models are gated and access is granted only to a select few. Hermes Agent now exposes MoA presets as virtual models, giving you capabilities beyond the publicly available frontier: 8% higher than Opus 4.8 and 11% higher than GPT 5.5 on our upcoming benchmark.

330 612 6K 1.8M 3K

Hermes now supports this! Game changer!!!!

Pavel Durov @durov

3 weeks ago

We now support rich formatting for all chatbots. Tables, nested lists, inline media, formulas, headers and more — right in Telegram messages. 🔨 Start building! Docs: core.telegram.org/bots/api#rich-…

489 559 8K 936K 2K

We're excited to join forces with @SpaceX to advance the frontier of useful AI. Expect significant improvements to Cursor soon.

SpaceX @SpaceX

3 weeks ago

SpaceX has exercised the option to acquire @cursor_ai in an all-stock transaction with the goal of building the world’s most useful AI models. For the past few months, SpaceXAI has been jointly training a model with Cursor, which will be released in Cursor and Grok Build soon.

2K 4K 37K 26.0M 3K

🤯 GLM-5.2 is here — built for long-horizon coding and agentic tasks, now with a solid 1M-token context. The strongest open-source coding model yet! Available now on Ollama's cloud, hosted in the US on the latest @NVIDIAAI Blackwell datacenter GPUs. Privacy policy and zero data retention apply, as always. Try it 👇 Claude Code: ollama launch claude --model glm-5.2:cloud Codex App: ollama launch codex-app --model glm-5.2:cloud Hermes Agent: ollama launch hermes --model glm-5.2:cloud Chat: ollama run glm-5.2:cloud More integrations and information in the model page 🧵

Z.ai @Zai_org

3 weeks ago

Introducing GLM-5.2: Frontier Intelligence, Open Weights - Significant improvements in coding and agentic tasks - Strong long-horizon capabilities with a 1M context window - Two levels of reasoning effort: GLM-5.2 (max) pushes the limits, while GLM-5.2 (high) strikes a strong

707 2K 13K 7.5M 5K

As a first blog post, I'm writing about how @polar_sh is steering away from Tailwind to StyleX in an effort to futureproof our frontend. Making it impossible to express an off-brand decision. If a value isn't a design decision, it should not typecheck. What do you want to know?

Birk Jernström @birk

3 weeks ago

Going to be a lot more public and vocal about building @polar_sh and behind the scenes. Anything people want to know?

27 1 189 60K 16

@emilwidlund @polar_sh Fun fact: @linear's internal component library is also called Orbit, and we’re also migrating to Stylex, from styled-components in our case.

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Claude models is not affected. We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible. Read our full statement: anthropic.com/news/fable-myt…

Den wichtigsten Satz gestern auf dem FDP-Parteitag sagte Martin Hagen, der neue Generalsekretär: „Mir ist lieber, wir begeistern die 20 bis 25 Prozent, die sich in Deutschland vorstellen können, die FDP zu wählen, als dass wir versuchen, es den anderen 75 bis 80 Prozent irgendwie recht zu machen, die uns am Ende sowieso doof finden.“ focus.de/politik/lieber…

Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities - Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas - MiniMax Sparse Attention scales context to 1M - Natively Multimodal from Step Zero API: platform.minimax.io Token Plan: platform.minimax.io/subscribe/toke… 🚀New! MiniMax Code: code.minimax.io Weights & Tech Report in ~10 Days

Absolut super Kommentar 🥳

Ulf Poschardt @ulfposh

a month ago

thomas berbner gehört zu dem 1% im ÖRR, der wirklich außerhalb der ÖRR blase lebt und denkt. formidabler kommentar.

373 2K 13K 427K 2K

I can't believe how fun building a company is right now is. The weird part is it doesn't feel like work anymore. New AI models/tools/repos keep coming out making the impossible possible. My ONLY anxiety is making sure I don't waste this precious moment and keeping up with the updates of all the new tools/AI models. AI is creating the greatest platform shift of all time. And, I've learned to never let a good platform shift go to waste. I was living in Silicon Valley around for the mobile era. I remember the feeling of "you can just build an app and put it in the store and people find it." That lasted maybe 4 years before the gold rush ended and distribution got hard. I'm getting that same feeling right now but bigger. The difference is I'm older, I know what a window looks like, and I know they close. I love building right now. Maybe you do too. Trying not to take it for granted. I'm excited for Monday. Can't wait. My partners and I are up at midnight most nights now sending each other screenshots saying "look what this can do." Nobody asks anyone to do this. We just can't stop. Something drops, someone builds something with it in 2 hours, and the group chat goes off. It feels like we're getting away with something. Some of the greatest companies of the next decade will be started in 2026. I'm sure of it. And it'll be fun. I feel like a kid again. Genuinely giddy.

Claude Opus 4.8 is insane. Nothing will be the same after this model. Anthropic should not have released something this dangerous.

Deutschland taumelt völlig unwissend in den wirtschaftlichen Niedergang. Selbst in meinem engsten Umfeld fehlt oft jedes Bewusstsein für die harten wirtschaftlichen Fakten. Viele sind völlig geschockt, wenn man ihnen die explodierenden Sozialkosten, den Wegfall industrieller Wertschöpfung, die Verlagerung von Produktion, den Stellenabbau, die immer weiter steigende Staatsquote, den China-Speed oder die Tech-Dominanz der Amerikaner erklärt. Dabei ist das alles seit Jahren alles sichtbar und bekannt. Das Land verliert seine wirtschaftliche Substanz, aber große Teile der Gesellschaft merken es wirklich erst, wenn es zu spät ist. Wenn ich zum Beispiel in großer Runde erkläre, dass wir in Deutschland inzwischen fast jeden Tag irgendwo ein Industriewerk verlieren, herrscht plötzlich Stille. „Oh, das wusste ich nicht. Ist das wirklich so? Wow.“ Schon allein deswegen hat der öffentlich-rechtliche Rundfunk, der eigentlich den Auftrag hat, hier zu informieren, völlig versagt. Es gibt aber auch ein totales Desinteresse an der deutschen Wirtschaft und ihren Problemen im überwiegenden Teil der deutschen Bevölkerung.

‼️ Die Partei des Wettbewerbs hat sich entschieden: Neuer Bundesvorsitzender der FDP ist @KubickiWo. Herzlichen Glückwunsch zur Wahl! #bpt26

Diese Grafik zeigt wie der gierige Staat unsere Mittelschicht zerstört hat. In den Jahren des deutschen Wirtschaftswunders bezahlten gut verdienende Arbeitnehmer einen Steuersatz von etwa 25 %. Familien konnten sich Häuser leisten. Ein Einkommen genügte. Das Fundament für eine stabile Gesellschaft wurde gelegt. Der Spitzensteuersatz in Höhe von 53 % griff inflationsbereinigt erst ab 1.072.000 €. Heute kennt das Steuersystem keine Mittelschicht mehr. Wenn Du eine guten Job hast, kommt sofort der Staat und verlangt 42 %. Vermögensaufbau wird unmöglich. Das eigene Haus wird zum fernen Traum. Linke Politik hat ein Steuersystem erschaffen, das die Bürger knechtet. Es ist hochgradig ungerecht.

Anthropic just shipped "dynamic workflows" and it's a much bigger deal than the Opus 4.8 release IMO here's what it actually means for your work: before, Claude worked like one person doing one task at a time. > you ask > it does the thing > you check it > you ask for the next thing. helpful, but it's still one worker and you're the manager watching over it dynamic workflows flips that. you describe a big goal, and Claude becomes the manager. it writes the plan, spins up hundreds of mini-Claudes working at the same time, has them check and challenge each other's work, and hands you back one finished result think of it like building a house before, you'd hired one handyman. you give him one task at a time, you watch him work, and if you want a whole house you're personally managing every nail now you've hired a full general contractor. you say "build the house, here's the inspection it has to pass," you approve the blueprint, and you leave. he brings in a whole crew working at once, runs inspectors who tear apart anything that's off, and you come back to a finished house that already passed inspection here's what that unlocks. the jobs that used to be too big to hand to AI: • move your entire app from one programming language to another while you sleep. someone just did exactly this on a massive project (750,000 lines of code) in 11 days with 99.8% of the tests still passing. that used to take a team months • run a full security sweep across your whole product at once, finding the weak spots and double-checking each one is real before it tells you • point it at an old, messy project nobody understands anymore and have it map out and explain the entire thing the wow factor is the scale + the walk-away. you kick off something that would've taken weeks, go to sleep, and it's still running and checking itself. if it gets interrupted, it picks up right where it left off. you're running a full on project now, not babysitting a little chatbot anymore you wanna reach for this when a task is: > too big to do in one shot > made of lots of independent pieces that can run at the same time > clear about what "done" means (like "all the tests pass") if it's small or fuzzy, a normal session is better and way cheaper. fair warning, this gobbles way more tokens than a regular session, so save it for the big jobs tldr: you used to hand Claude a task. now you hand it a project

cat @_catwu

a month ago

Excited to share our most powerful new Claude Code feature: dynamic workflows! Mention "workflow" in a prompt and Claude will dynamically create an orchestration plan that it strictly follows, allowing you to confidently trust that every stage happens in the right order even

354 829 8K 1.7M 6K