jedisct1

Swival Audit feature gives really good results even with small, open models: https://swival.dev/pages/audit.html

This is not an issue with Claude, but with Claude Code.

With swival.dev you can use long sessions without ever doing any manual compaction or reset, even with Claude models.

https://swival.dev is an AI agent that was designed for open models since day 1.

"This browser doesn't support WebAssembly JSPI, which Firefox WASM needs to run."

What a waste of tokens.

Not opensource.

I build a Signal client for Swival: https://github.com/swival/nbclaw

And I use it daily for simple tasks.

GLM-5.2 is a great model!

But it already works really well with existing harnesses, I'm not sure why a dedicated one is needed?

I use it with https://swival.dev and everything works perfectly, no tool calling issues and it works fine with long sessions.

Monetization Gateway for Bunny CDN: https://github.com/dip-proto/x402

and for Fastly: https://github.com/dip-proto/x402-fastly

doh-server supports DoH but also ODoH: https://github.com/DNSCrypt/doh-server

This.

FOr Mac users: https://huggingface.co/collections/jedisct1/qwen-agentworld

Exactly!

I once had some pretty serious DoS attack, but fortunately, I didn't had to pay more than what was pre-paid.

I got significant improvement on code quality (so much that it has become a no brainer for important tasks such as planning) simply by adding the --self-review flag to swival: https://swival.dev/pages/reviews.html

Two instances of the same model, a producer and a reviewer, and the loops doesn't end until everybody's happy.

Also https://00f.net/2018/11/25/webassembly-doesnt-make-unsafe-la...

Misleading title. This is about Claude for Apple Foundation Models, not about Apple Foundation Models

There's https://github.com/ofalkenberg/zig-wasi-components

The best usage example of WASI I know of is the Zig compiler: https://ziglang.org/news/goodbye-cpp/

Because they didn't store data before? Don't be so naive.

A lot of things can be done with local models.

When agents don’t encrypt secrets, MCP servers help prevent users from handing their API tokens to AI providers or intermediaries such as Cloudflare and Akamai.

It's probably good, and the best for Deepseek models, but do we really need one harness per model?

You can get a taste of this today yourself with Swival /audit command and the security scanner is going to get even closer soon: https://medium.com/@swival/ai-vulnerability-scanning-needs-a...

Have you tried Swival /audit? https://swival.dev/pages/audit.html

Microsoft’s GitHub was compromised when a Microsoft developer using Microsoft VSCode installed a rogue extension from Microsoft’s VSCode extension library, which is moderated and hosted by Microsoft.

Interesting!

The https://swival.dev harness already has retry nudges, step enforcement, error recovery, context awareness, etc. to try to support small models as much as possible.

Curious to see how it compares with forge, and if both could be combined.

Depends on the language and harness, I guess.

It works really well for me, at least for Python and JavaScript, with swival.dev as a harness.

Swival found many more vulnerabilities without Mythos https://github.com/swival/security-audits

Finding vulnerabilities everywhere doesn't need any skills and more, nor Mythos.

See https://github.com/Swival/security-audits/ for examples, which are automated security audits just made with swival.dev /audit command, and includes audits of large code bases such as the entire OpenBSD base system.

Agent Metaskills: https://swival.dev/pages/metaskills.html