Tommi Somersuo
@tommis.fi
AI Tech Lead @ Solita.fi | Agentic engineering, architecture & governance.
Offline: Runner, angler, RES/SRA, Tibetan Terrier owner.
Offline: Runner, angler, RES/SRA, Tibetan Terrier owner.
Reposted by Tommi Somersuo
"AI sovereignty doesn’t necessarily imply total control of your AI stack. It holds the promise of having more localized security and privacy, better adherence to local jurisprudence, more dependable service, and more culturally specific outputs." https://bit.ly/4AoHUx2
AI Sovereignty: Bargaining with Big Tech and the Promise of Full Stack Open Source AI
bit.ly
September 22, 2026 at 6:39 PM
"AI sovereignty doesn’t necessarily imply total control of your AI stack. It holds the promise of having more localized security and privacy, better adherence to local jurisprudence, more dependable service, and more culturally specific outputs." https://bit.ly/4AoHUx2
Reposted by Tommi Somersuo
Fantastic benchmark of Jev vs its "clones" from HuggingFace: DecisionIndex huggingface.co/spaces/multimo...
September 22, 2026 at 7:20 AM
Fantastic benchmark of Jev vs its "clones" from HuggingFace: DecisionIndex huggingface.co/spaces/multimo...
Reposted by Tommi Somersuo
I was asked by Congressional staff to share my views on open models performance, adoption, and competition vis a vis China. Here’s my briefing with all our latest data and predictions for what comes next.
www.interconnects.ai/p/the-curren...
www.interconnects.ai/p/the-curren...
The current balance of power in open models
The expanded form of a testimony I prepared for Congress.
www.interconnects.ai
September 21, 2026 at 12:01 PM
I was asked by Congressional staff to share my views on open models performance, adoption, and competition vis a vis China. Here’s my briefing with all our latest data and predictions for what comes next.
www.interconnects.ai/p/the-curren...
www.interconnects.ai/p/the-curren...
Reposted by Tommi Somersuo
[ Love this! ]
"Build the smallest brain that plays well. Train a neural network, describe it in a manifest, and submit both. 16 KiB is a whole weight class, so the question is not how big a model you can train but how little it takes."
tinybrains.dev
"Build the smallest brain that plays well. Train a neural network, describe it in a manifest, and submit both. 16 KiB is a whole weight class, so the question is not how big a model you can train but how little it takes."
tinybrains.dev
TinyBrains — build the smallest brain that plays well
A ranked ladder for small neural networks that play strategy games. Train a model, write an adapter, upload two files, and get measured into a weight class from 8 KiB up.
tinybrains.dev
September 21, 2026 at 10:44 AM
[ Love this! ]
"Build the smallest brain that plays well. Train a neural network, describe it in a manifest, and submit both. 16 KiB is a whole weight class, so the question is not how big a model you can train but how little it takes."
tinybrains.dev
"Build the smallest brain that plays well. Train a neural network, describe it in a manifest, and submit both. 16 KiB is a whole weight class, so the question is not how big a model you can train but how little it takes."
tinybrains.dev
Reposted by Tommi Somersuo
📜 introducing ossrules.md
- learn how leading OSS projects use agent rule files
- discover skill files that projects actually use
- level up your own projects and workflows
I've personally learned a lot just exploring
- learn how leading OSS projects use agent rule files
- discover skill files that projects actually use
- level up your own projects and workflows
I've personally learned a lot just exploring
September 18, 2026 at 8:49 PM
📜 introducing ossrules.md
- learn how leading OSS projects use agent rule files
- discover skill files that projects actually use
- level up your own projects and workflows
I've personally learned a lot just exploring
- learn how leading OSS projects use agent rule files
- discover skill files that projects actually use
- level up your own projects and workflows
I've personally learned a lot just exploring
Claude Code 2.1.277: Added AGENTS.md support: in a project with no CLAUDE.md, Claude Code reads AGENTS.md instead; change it under "Project instructions" in /config (not yet on Bedrock, Vertex or Foundry)
claude-code/CHANGELOG.md at main · anthropics/claude-code
Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflo...
github.com
September 19, 2026 at 1:01 AM
prismml.com/news/bonsai-...: Introducing Bonsai 2 27B: Near-Lossless Compression in a 9x Smaller Footprint
September 18, 2026 at 6:41 AM
prismml.com/news/bonsai-...: Introducing Bonsai 2 27B: Near-Lossless Compression in a 9x Smaller Footprint
Reposted by Tommi Somersuo
Here's what OpenAI's agentic software factory looks like, today.
It's surely not token efficient, but eg agents "babysitting" releases to prod is something new, and an interesting approach.
Details: newsletter.pragmaticengineer.com/p/openai-sof...
It's surely not token efficient, but eg agents "babysitting" releases to prod is something new, and an interesting approach.
Details: newsletter.pragmaticengineer.com/p/openai-sof...
September 15, 2026 at 7:38 PM
Here's what OpenAI's agentic software factory looks like, today.
It's surely not token efficient, but eg agents "babysitting" releases to prod is something new, and an interesting approach.
Details: newsletter.pragmaticengineer.com/p/openai-sof...
It's surely not token efficient, but eg agents "babysitting" releases to prod is something new, and an interesting approach.
Details: newsletter.pragmaticengineer.com/p/openai-sof...
Reposted by Tommi Somersuo
Jev: Fable-level model that doesn’t charge for output tokens because they’re too cheap to meter
it’s not general though, it only makes decisions, doesn’t generate text, but input tokens are measured by the billion ($42/btok)
typesafe.ai/blog/introdu...
it’s not general though, it only makes decisions, doesn’t generate text, but input tokens are measured by the billion ($42/btok)
typesafe.ai/blog/introdu...
September 15, 2026 at 8:13 PM
Jev: Fable-level model that doesn’t charge for output tokens because they’re too cheap to meter
it’s not general though, it only makes decisions, doesn’t generate text, but input tokens are measured by the billion ($42/btok)
typesafe.ai/blog/introdu...
it’s not general though, it only makes decisions, doesn’t generate text, but input tokens are measured by the billion ($42/btok)
typesafe.ai/blog/introdu...