Sabitlenmiş Tweet
RAG is overkill for small doc sets. If you have a few hundred markdown files, you don’t need embeddings or a vector DB - you need grep.
Built: memexfs - WASM grep/read/ls for Node, zero deps.
github.com/aspelund/memex…
English
Mattias Aspelund
34.8K posts
Mattias Aspelund
@aspelund
Techie, Co-Founder Awolve
Stockholm, Sweden Katılım Mart 2009
1.2K Takip Edilen1.2K Takipçiler
@rickasaurus Its all fun until it makes something different or re-seeds a database without being asked to :)
And honestly, it doesn’t take more time to ask it to keep a deploy script up to date.
CI/CD on merge to specific branches, that might be gone though.
English
@FakePsyho Really? I saw a huge leap in capabilities with 5.5 too.
English
Improve crosswalk safety with @ultralytics YOLO26! 🚸
Detect pedestrians, vehicles, and high-risk interactions
in real-time to support safer crossings, smarter alerts,
and improved traffic monitoring.
#RoadSafety #MachineLearning #Research
English
@gabriberton Even if it would work optimizing your harness for something that clearly will eventually be removed is a waste of time.
English
Don't fall for this. It does not work
I wanted to write a series of posts about related stuff for almost a year now, I promise the time is almost here
International Cyber Digest@IntCyberDigest
This is absurdly clever: there is a way you can cut Fable 5 costs by up to ~70%. Just turn Claude Code context into an image and make Fable OCR them. 😂 github.com/teamchong/pxpi…
English
@jun_song I used it for several tasks today without it getting routed - one was around doing root cause analysis for a defect with lots of code and infra log analysis where I almost expected it to route. But it didn’t.
English
Please @OpenAI release the new models today! I want to see the real magic of the art. Of the possible!
English
Letting Fable plan and outline a RCA for a very very weird corner case and then letting GPT-5.5 xhigh execute it feels nothing less than magical!
English
Seems like Sonnet 5 is best (and cheapest) for high precision tasks with well defined boundaries. I.e. when you want repetitive knowledge tasks automated where output quality is important.
English
@OliverMolander Thw whole sovereign debate is mostly ridiculous. The non-EU dependency graph in tech is endless. At best, we can get a better hand for negotiating.
English
You can only be truly sovereign by building globally leading technology that others count on.
Seb Johnson@SebJohnsonUK
Europe has a massive obsession with tech sovereignty. @OliverMolander thinks that mindset misses how to actually win. Building safe local tech doesn't give you leverage. You only win when you build something so critical that the rest of the world panics if you take it away.
English
@trq212 If you are excited, imagine how effing excited we are. :P
English
Have seen some questions about the updated classifiers and wanted to clarify.
As with the original classifiers, a small fraction of routine coding and debugging tasks will be flagged and fall back to Opus.
We're excited for guys to get access back tomorrow.
Anthropic@AnthropicAI
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…
English
@elonmusk Did you mostly but Cursor for the data, models or the knowhow?
English
Grok 4.5, based on our 1.5T V9 foundation model, with Cursor data added in supplemental training, is now in private beta at SpaceX & Tesla. Early evals show performance close to, perhaps exceeding Opus.
RL is continuing to significantly improve the model, and the Grok Build harness gets better every day.
Nice work by all those involved!
Completely trained from scratch new models will be released by @SpaceX every month this year.
English
@OliverMolander Lets do it! And given our track record from building JAS, and our digital track record, we should at least be able to be on par with the Chinese.
English
Idea:
Let's create a world-leading Nordic AI research lab.
There's an insane amount of top AI research talent from the Nordics.
I have many friends who work or have worked in top positions at e.g. Deepmind, Anthropic, OpenAI.
No region can match the Nordics in trust.
English
@ivanfioravanti Eventually stopping a scalable resource such intelligence will cause the same kind of pushback as other isolationist measures. Might require some negotiating, but it will be solved.
English
What will happen if GPT 5.6 won’t be available worldwide?
Rest of the world will go all in on GLM, MiniMax, Kimi, DeepSeek, StepFun, Mistral, HuggingFace and we’ll live happily ever after in our worldwide community.
Sam Altman@sama
@nadzi_mouad working hard for worldwide
English
@Hesamation Being fun at parties is an overrated skill for people changing the world.
English
I bet he’s fun at parties.
Keith@gnukeith
This guy has been fearmongering professionally since DAY ONE
English
I am so deeply disappointed by the US gov. It was apparently a mistake to think this administration had a good sense to support American AI models.
However, I also put more blame on the cult of Anthropic & Dario Amodei. All that fearmongering both achieved its goal & backfired!
TravisGood@IridiumEagle
The US gov is playing a dangerous game. If Chinese models soundly surpass Opus 4.8 while Mythos and GPT 5.6 are banned, the narrative around 'distillation' will collapse, and US AI valuations and adoption will face an existential crisis.
English
@yubai01 And congrats on finally getting into a good naming game. :)
English
GPT-5.6 is a strong model. We are already heavily relying on it in Codex, can't wait to make it available to everyone.
Other highlights:
* Three new models lined up, Terra/Luna are closer to the flagship model than previous generations of mini/nanos -- no longer "mini models" per se
* Max reasoning mode -- thinks longer
* Ultra mode -- uses subagents automatically
OpenAI@OpenAI
Sol is our new flagship and a step function better than GPT-5.5. Terra delivers performance competitive to GPT-5.5 at 2x lower cost. Luna is our most cost-efficient model, delivering strong capability at our lowest cost. Together, the GPT-5.6 family gives people and developers more choice in how they balance intelligence, speed, and cost.
English
@gabriel1 If some patterns are established, like pickers and similar, dynamically building ”this is what needs your attention” with predictable patterns comes naturally. And not years away,
English
there are zero dynamic interfaces or apps that tailor themselves to the user that i know of
can someone enlighten me why ai suddenly changes this
just moving a single button will confuse an entire enterprise, when do we learn that feature discovery is the worlds hardest problem
English
Mattias Aspelund retweetledi
@palu1161 @lindgren_p @OlssonKarin @CarlOskar För stort fokus på oppositionen och de mest polariserande frågorna. Undrar vad som hade hänt om de konstant bara pratat om god vardag för medelsvensson istället.
Svenska
@lindgren_p @OlssonKarin @CarlOskar Fascinerad över hur Tidö lider mot så gigantiskt nederlag, trots ganska bra realpolitik. Är uppenbarligen enormt ”un-likable” 😬
Kan ha att göra med polarisering, eller grabbig ton som alienerar kvinnor, eller annat. Men fascinerande! 🤷♂️
Svenska
Fascinerande det där med att man inte kan veta hur andra uppfattar världen. Kanske ser någon gult så som jag ser färgen grönt.
@OlssonKarin ser att @CarlOskar blev ”utskälld” och ”fick sina fiskar varma” när jag ser att hon blev totalt överkörd och tillintetgjord.
Carl-Oskar Bohlin@CarlOskar
Jag tror och hoppas att detta tjänar som en klargörande diskussion och möjligen illustrerar den ett eller två förlegade perspektiv som finns hos somliga representanter för traditionella medier. Jag tror att demokratin mår bra av att politiker kan kommunicera med väljare direkt och i egna kanaler.
Svenska
GPT-5.6 Sol preview — it's a good model:
OpenAI@OpenAI
Introducing a limited preview of GPT-5.6 Sol, our next generation frontier model, as well as GPT-5.6 Terra, a balanced model for efficient, everyday work, and GPT-5.6 Luna, a fast and affordable model for high-volume work. openai.com/index/previewi…
English