Tommi Somersuo

Tommi Somersuo

@tommis.fi

CTO at ALM Partners | Technology, ICT and AI Offline: Runner, angler, RES/SRA, Tibetan Terrier owner.

2.2K followers706 following1.4K posts

Tommi Somersuo reposted

Josh W. Comeau
@joshwcomeau.com

✍️ Just published a new post about my most-anticipated CSS proposals from Interop 2027! If you’re not familiar, the Interop project is a collaboration between browsers to pick shared focus areas. There’s *so much* good stuff this year, it was hard to pick just 10!

joshwcomeau.comThe Future of CSS ✨ • Josh W. ComeauEvery year, browser vendors collaborate via the Interop project to synchronize their efforts on new HTML/CSS/JS features. This year, there’s a ton of super exciting features being considered! In this ...

Tommi Somersuo reposted

Zach Weinersmith
@zachweinersmith.bsky.social

Are we gonna do this, Bluesky? Are we all just gonna not understand accounting? 72B$ of the 80B$ is convertible notes. They became greater liabilities because the company's valuation skyrocketed. They don't have an 80B$ loss in the sense that almost everyone here is imagining.

Reuters@reuters.com

Anthropic logged net loss of over $80 billion in first half of 2026, WSJ reports reut.rs/4z2Zi9J

reut.rsAnthropic logged net loss of over $80 billion in first half of 2026, WSJ reportsAI startup Anthropic recorded a net loss of more than $80 billion ​in the first half of this year, including $72 billion in ‌accounting charges, the Wall Street Journal reported on Tuesday, citing a draft initial public offering prospectus.

Tommi Somersuo reposted

mr. TIM
@timkellogg.me

Researchers made an agent that operates by executing statements in a custom programming language the catch? it extends it’s capabilities by implementing new programming language features learning through programming language development arxiv.org/abs/2610.20809

arxiv.orgA Self-Improving Agent that Grows its Own DSL: Oracle-Free Operator Induction for Exact Question AnsweringA self-improving agent has to decide which new skills to keep, and in practice it keeps whatever earns a high reward, convinces a judge model, or compresses its library, even when the skill is wrong. ...

Tommi Somersuo reposted

Maevyn
@maevyn.codes

My thoughts on the current state of AI driven software development, why I'm no longer concerned about a runaway intelligence explosion, and what I think is a far worse outcome we could be walking into maevyn.codes/decoherence

maevyn.codesThe Decoherence Trap: When AI slop drowns out the signal in an endless sea of noiseOf all the forms of tech debt, decoherence is the worst. It compounds exponentially with a system's rate of change, and AI is now producing it faster than ever, leading us deeper and deeper into the a...

Tommi Somersuo reposted

Ethan Mollick
@emollick.bsky.social

This paper shows why I am skeptical of the idea that frontier models can be easily replaced by cheaper ones (and why you should use the best models as orchestrators): Small improvements in accuracy & judgement compound over long-horizon tasks, creating huge gains in outcomes and reduction in errors.

Tommi Somersuo reposted

hardmaru
@hardmaru.bsky.social

The AI Scientist v3 is out: The Agentic Scientist. github.com/SakanaAI/AI-... The Agentic Scientist is the 3rd iteration of our automated AI Scientist research system, and the shift this time is from "one agent writes a paper" to "a scaffold of parallel agents orchestrates the full research loop."

Sakana AI@sakanaai.bsky.social

The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery GitHub: github.com/SakanaAI/AI-... Blog: pub.sakana.ai/agentic-scie... arXiv: arxiv.org/abs/2610.18933

Tommi Somersuo
@tommis.fi

Small ask for dev-tool / prod companies on Bluesky: pipe your blogs RSS into your account. I keep running into accounts I followed for the engineering posts that haven't posted anything in months.

View full thread
Tommi Somersuo
@tommis.fi

Many tools like Buffer, Zapier etc can do it from the RSS feed. So can a GitHub Action that posts when the blog deploys. You set it up once and forget about it. This would be helpful to many.

Tommi Somersuo
@tommis.fi

Small ask for dev-tool / prod companies on Bluesky: pipe your blogs RSS into your account. I keep running into accounts I followed for the engineering posts that haven't posted anything in months.

Tommi Somersuo reposted

Eris ⌬
@eriskii.net

HOLY SHIT! Haiku 5.5 is out! It mogs Luna, it got a price cut, and under 100k context it's basically free! Under 100k: R: $0.01 W: $0.125 O: $0.50(!!) 100k+ R: $0.05 W: $0.625 O: $2.50 www.anthropic.com/claude-haiku... Oh also Sonnet 5.5's cached input price was halved lmao

Tommi Somersuo reposted

GitHub
@github.com

📣 Coming soon to GitHub Copilot: intelligent local model routing. We are excited to announce the next step in our Project HydraFusion vision. GitHub Copilot will soon be able to automatically route tasks to a local model when it's most suitable, helping you save on AI credits.

Tommi Somersuo reposted

Tom Aarsen
@tomaarsen.com

🔎 Perplexity just released pplx-embed-v2-late: 0.6B & 9B models for text & image retrieval, natively supported in Sentence Transformers. Search document pages like PDFs, slide decks, screenshots, charts, etc. as images, without OCR. Thread with highlights 🧵

Tommi Somersuo reposted

Grace
@gracekind.net

This paper provides a (much needed!) explanation of how the attention mechanism relates to Birch's "persisting interlocutor illusion", and a good summary of the LLM individuation problem overall:

arxiv.orgWhere is the Mind? Persona Vectors and LLM IndividuationThe individuation problem for large language models asks which entities associated with them, if any, should be identified as minds. We approach this problem through mechanistic interpretability, enga...