frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Show HN: Submilli – runtime with semantic permissions for agents that write code

https://github.com/submilli/submilli-runtime
3•somdoron•53m ago

Comments

somdoron•52m ago
Submilli is an open-source runtime for business agents that use programmatic-tool-calling or code-mode.

Submilli’s permissions model works from the inside out. Submilli has packages, like every other language. The package defines the permission vocabulary for a capability. For example, a refund capability comes with a payload that includes a customer ID and an amount. The interpreter checks, in real time, if that capability was granted to the agent, and denies or allows accordingly. That allows writing rules like “Allow a refund up to $500, only for customer 123”.

Submilli compiles TypeScript to WebAssembly, comes with ECMA-262 and a standard library with a restricted API to the outside world, which is also governed by the above permission model.

In race with U.S., China struggles to recruit foreign AI researchers

https://www.nytimes.com/2026/10/06/science/china-ai-research-recruitment.html
1•ilamont•11s ago•0 comments

Apple's App Icon HOA

https://blog.jim-nielsen.com/2026/app-icon-hoa/
1•Brajeshwar•1m ago•0 comments

NTS: Authenticated Time at Meta

https://engineering.fb.com/2026/10/06/production-engineering/nts-authenticated-time-at-meta/
1•acossta•2m ago•0 comments

Treg (OpenRouter for Tools)

https://github.com/superdesigndev/treg
1•trollied•2m ago•0 comments

Software developers are not okay

https://www.baldurbjarnason.com/2026/05-software-developers-are-not-okay/#software-developers-are...
2•jalev•3m ago•0 comments

You have to be able to reason about the verification

https://www.brandons.me/blog/you-have-to-be-able-to-reason-about-the-verification
2•brundolf•4m ago•0 comments

k-space (a.k.a. Reciprocal Space)

https://www.bohrium.com/en/sciencepedia/feynman/keyword/k_space
1•peter_d_sherman•4m ago•0 comments

Python's two modules for random numbers

https://lwn.net/SubscriberLink/1097468/63d60212ee6a2e98/
1•chmaynard•5m ago•0 comments

Using Conditional Value at Risk to Illustrate Two Risk Paradigms in Golf

https://golfcoursewiki.substack.com/p/accuracy-outcomes-and-the-good-miss
1•scoofy•5m ago•0 comments

Show HN: SPrav Job AI Client-side career tool with WebGPU and direct ATS queries

https://sprav-jobai.vercel.app/
1•SVSPraveen•5m ago•0 comments

Routing PCBs on a hex cell grid

https://substack.com/sign-in
1•jamesbowman•7m ago•0 comments

AI is now capable of developing its own inference hardware

https://github.com/FeSens/openTPU
4•fsbonetto•9m ago•2 comments

Hark Pro – Personal AI that can use the web and get things done for you

https://hark.com
1•nderjung•9m ago•0 comments

Bolt Playground

https://bolt.troymoder.dev/
1•arunc•10m ago•0 comments

Flatpak from the CLI Sucks

https://kowalski7cc.xyz/blog/flatpak-from-the-cli-sucks/
1•ibobev•10m ago•0 comments

Process vs. Talent (Workflow vs. Model)

https://jaksa.me/blog/2026-10-03-portraits
1•jaksa•10m ago•1 comments

I gave an AI agent one bug 11 times. 6 commits had files it never knew about

https://junayed.com/blog/same-bug-eleven-times/
1•mjmizan•10m ago•0 comments

Google enters 3.6-GW power deal with Constellation Energy

https://www.reuters.com/business/energy/google-enters-massive-36-gw-power-deal-with-constellation...
1•onemoresoop•11m ago•0 comments

Scrimshaw Jukebox

https://simonwillison.net/2026/Oct/6/scrimshaw-jukebox/
1•rdslw•11m ago•0 comments

Animated Git Cheat Sheet

https://initialcommit.com/learn/git/animated-cheat-sheet
1•ibobev•11m ago•0 comments

Hacking the Go compiler to efficiently map IPv4 to IPv6

https://vincent.bernat.ch/en/blog/2026-go-netip-addrto6
1•ibobev•12m ago•0 comments

Google AI Edge Gallery

https://developers.google.com/edge/gallery
1•tosh•13m ago•0 comments

Hopf Spherical Quantization (& Compression)

https://github.com/meridionalissoftware/hscq
1•binge_eating•17m ago•0 comments

Chrome sneakily downloads Google Meet code

https://webrtchacks.com/chrome-neural-echo-cancellation/
1•slac•18m ago•1 comments

How Machines Learned Precision

https://glinscott.github.io/how-machines-learned-precision/
1•glinscott•18m ago•1 comments

UBS CEO warns 'hard measures' are needed to tackle French debt crisis

https://www.cnbc.com/2026/10/06/france-debt-crisis-ubs-ermotti.html
1•Betelbuddy•18m ago•0 comments

Escaping the Tyranny of Worktrees

https://delta.dev/blog/escaping-the-tyranny-of-worktrees
1•ilreb•18m ago•0 comments

Proving Rust Web Application Correctness with Lean 4

https://medium.com/@Koukyosyumei/proving-rust-web-application-correctness-with-lean-4-8889583f1e15
1•syumei•19m ago•0 comments

Show HN: Capo-S3 – Modern S3 client for Python

https://pypi.org/project/capo-s3/
2•karpetrosyan•19m ago•0 comments

OpenAI is adding text watermarking in ChatGPT and Codex

https://www.theverge.com/ai-artificial-intelligence/1004880/openai-chatgpt-text-watermarks-eu-ai-act
1•thm•19m ago•0 comments