frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Show HN: Silta: family assistant on Matrix with continuity across context limits

https://github.com/dmitry-markin/silta
3•dmitry-markin•1h ago
Silta is a self-hosted assistant serving multiple users from a single Matrix account, used by me, family, and close friends for the last three weeks. It runs on Claude Code with each person using their own Claude subscription.

Silta answers everyday questions, conducts web research (presenting the results as PDFs), performs scheduled tasks to monitor/research something and notify you if needed.

It uses a deliberate handoff & compaction procedure, writing the compaction summary itself with a custom prompt, so the thread of the conversation and its picture of the user is preserved across context limits. As it turned out, the assistant accessible in a continuous chat (contrary to ChatGPT / Claude Web sessions) builds up better context about you and your tasks, and provides more relevant answers and help.

It runs on Claude Code as the only allowed way to use a personal Claude subscription. Using Claude Code also spares me developing a harness, at the cost of tracking changes as new versions of Claude Code are released. The assistant itself (the daemon, the channel plugin, the supervisor, and the deployment) is open-source, while the model and the harness are Anthropic's and proprietary.

Silta is deployed to a dedicated VM with separate Linux users per session and systemd hardening, protecting sessions from each other. Claude Code's bubblewrap sandbox is used to protect the Claude credential and the harness's state from the agent. This does not protect from prompt injections coming from search results and potentially leaking user data.

Matrix E2EE protects the conversation from a Matrix server operator, but everything in the context is of course sent to the model provider. Private content is never logged, and the operator is expected to respect the privacy and not read the session transcripts on disk. Additionally, it is recommended to use full-disk encryption and tell the users how their data is handled.

Silta only works in its workspace with files explicitly shared by users in the chat. Giving it access to a user's own computer or services like email is not planned.

Since Opus 5.5 has been released, it is possible to run a session on a $20 Claude Pro subscription. Before that, only Fable 5.1 was pleasant enough to talk to. When run on Anthropic API, a moderately used session extrapolates to $100-$150 per month, with each research run costing $6-$12.

Silta is built with Claude Code and Silta itself. The design, review, deployment, and testing are mine.

Effective altruism is this century's biggest idea

https://www.economist.com/leaders/2026/10/01/effective-altruism-is-this-centurys-biggest-idea
1•Anon84•51s ago•0 comments

Rowbird – Open-source SQL scheduler for alerts and reports

https://rowbird.dev/
1•fabioselau•1m ago•0 comments

Show HN: Let customers automate tasks inside your SaaS

https://embedworkflow.com/
1•ewf•1m ago•0 comments

Zenkai: A Zig App Launcher That Starts in 20ms Sometimes

https://dayvster.com/blog/zenkai-a-zig-app-launcher-that-starts-in-20ms-sometimes/
1•ibobev•1m ago•0 comments

Circular relationships among AI firms

https://www.bis.org/publications/bulletin-137-circular-relationships-among-ai-firms
1•AnodicElegy•2m ago•1 comments

We linted 6,008 real database schemas. 45% of foreign keys have no index

https://blog.mcdview.dev/en/we-linted-6000-real-database-schemas/
1•PatuDev•2m ago•0 comments

Fast Blur with Animated Radius

https://aras-p.info/blog/2026/10/01/Fast-blur-with-animated-radius/
1•ibobev•2m ago•0 comments

Windows on AArch64 also provides for hot-patching, but much simpler than on x86

https://devblogs.microsoft.com/oldnewthing/20260930-00/?p=112744/
1•ibobev•3m ago•0 comments

Show HN: I made a technical trivia quiz app graded by AI

https://nadir.dev/
1•gmitrev•4m ago•0 comments

Memory Snapshots for CPython (Python Language Summit 2026)

https://blog.python.org/2026/09/language-summit-2026-memory-snapshots/
1•bucket2015•4m ago•0 comments

How to Wire Jev into Pi Coding Agent Without Losing Your Mind

https://grigio.org/how-to-wire-jev-into-pi-coding-agent-without-losing-your-mind/
2•grigio•5m ago•0 comments

Show HN: OIDC Inspector

https://contextiq.trango-compute.com
1•contextiq•5m ago•0 comments

OST to PST Converter Is Convert Exchange Outlook OST Files to Outlook PST

https://blog.perfectdatasolutions.com/ost-to-pst-converter-software-2/
1•tieanderson•5m ago•1 comments

Christa Gail Pike survives lethal injection attempt

https://apnews.com/article/christa-pike-execution-tennessee-f826a60582083b90ac7695ecb16afbdf
1•the_arun•6m ago•0 comments

An Overview of Differential Geometry

https://derangedmathematician.substack.com/p/an-overview-of-differential-geometry
1•chadcmulligan•6m ago•1 comments

Anyone else develop a system for open horizon tasks with minimum supervision

1•K0balt•6m ago•0 comments

R2 is partially down (slow)

https://www.pulsapi.com/services/r2
1•pinter69•7m ago•0 comments

McDonald's Is Reportedly Using AI to "Dynamically" Price Its Burgers

https://www.engadget.com/2272211/mcdonalds-is-reportedly-using-ai-to-dynamically-price-its-burgers/
1•speckx•7m ago•0 comments

Missouri issues new guidelines for Flock cameras, license plate readers

https://www.kmbc.com/article/missouri-flock-camera-license-plate-reader-guidelines/73968021
1•lenerdenator•9m ago•0 comments

Show HN: Omadesign 0.6.3 – Linux creative suite, cloud review, CJK IME

https://omadesign.app
1•michaelh_rley•10m ago•0 comments

We want you to build the next Git platform on Cloudflare

https://blog.cloudflare.com/next-git-platform-on-cloudflare/
1•tosh•10m ago•0 comments

Show HN: Nixm – AI chat that keeps no transcript (we measured the trade-off)

https://nixm.ai/
1•prousselet•10m ago•0 comments

Trump's AI rebrand causes 'unprecedented' demand for Slovenian website names

https://www.bbc.co.uk/news/articles/cqx2z23xj555o
3•timoth•11m ago•0 comments

Show HN: InferPilot – the vLLM doctor that tells you when fp8 won't help

https://github.com/poojithdevan4D/InferPilot
1•poojithdevan•12m ago•0 comments

Namespace – San Francisco/ New York / Zurich| Full-Time – Remote/Hybrid

https://namespace.so
1•hannahrsmith•12m ago•1 comments

Mitigating Memorization in LLMs

https://blog.janestreet.com/mitigating-memorization-in-llms/
1•torutofu•13m ago•0 comments

AWS reimagines the getting started experience

https://aws.amazon.com/blogs/aws/aws-reimagines-the-getting-started-experience/
2•aspittel•13m ago•0 comments

Is the Food Industry in Trouble?

https://www.foodpolitics.com/2026/09/is-the-food-industry-really-in-trouble/
1•speckx•13m ago•0 comments

The most praised thing I ever built was a dropdown

https://www.vinoth.net/littlethings
1•avinoth•14m ago•0 comments

Meta Uses A.I. Data Centers to Avoid Billions in Federal Taxes

https://www.nytimes.com/2026/09/30/technology/meta-ai-data-centers-taxes.html
3•gmays•14m ago•0 comments