SilenN

5 ms·

1. 1.▲
   by SilenN 19d ago

   "Anonymous aggregate PostHog product telemetry is enabled by default. It never includes prompts, traces, actions, observations, paths, model names, credentials, or raw customer content."
2. 2.▲
   by SilenN 20d ago

   We had this exact problem so we solved it for ourselves. Happy to help if you run into any issues. I have a /hmmm command I use for the second part that works reasonably well: "Stop using jargon and speak coherently. State it more
3. 3.▲
   by SilenN 20d ago

   Thank you!
4. 4.▲
   by SilenN 20d ago

   Thanks for the positivity tyre! If you look at our git history, we pivoted and only started building the gateway recently. Before that we were building research infrastructure that now powers the intelligence features we provide.
5. 5.▲
   by SilenN 20d ago

   Telemetry is off by default. PostHog is for usage analytics on the open source repo. Audit it if you're skeptical. We make money off enterprise licenses and hosting models. I have strong reason to suspect you either can't read or
6. 6.▲
   by SilenN 20d ago

   Yep! One interesting example is often Opus 5 on low reasoning ~= Opus 5 on high reasoning.
7. 7.▲
   by SilenN 20d ago

   See you soon
8. 8.▲
   by SilenN 20d ago

   Ans: we rarely switch, often times it's just a "switch to using this model for your agent"
9. 9.▲
   by SilenN 20d ago

   Also a hosted marketplace, not just BYOK
10. 10.▲
    by SilenN 20d ago

    That is another way to do. Or we can automatically figure out which models the subagents should be using for you. And update them as new models come out and the work your subagents do changes. More than one way to skin a cat.
11. 11.▲
    by SilenN 20d ago

    Yep! If you look at the commit history that's exactly what happened.
12. 12.▲
    by SilenN 20d ago

    The trick is to rarely switch, or switch at task boundaries. Often the conclusion of routing is actually "this one model is actually at the pareto front for this task, just use it always".
13. 13.▲
    Show HN: We built open OpenRouter that turns usage into a better model(github.com)

    222 points by SilenN 20d ago|47 comments
14. 14.▲
    by SilenN 2mo ago

    ^this There's two ways to functionally measure this, reconstruction fidelity (which we're able to get to 0.7 - 0.95), and downstream performance (which agrees on the best and worst models, but can sometimes disagrees on the middle
15. 15.▲
    by SilenN 2mo ago

    Exactly
16. 16.▲
    by SilenN 2mo ago

    Expensive, in the thousands. We have our own infra in house and are working on bringing these costs down
17. 17.▲
    by SilenN 2mo ago

    Technically 0 because a) it ingests your already existing traces and does an initial training run b) in the app we'll have pre-trained routers you can start with that will then learn over time
18. 18.▲
    by SilenN 2mo ago

    Fixed formatting which will help with readability. We do routing, distillation, and token compaction.
19. 19.▲
    by SilenN 2mo ago

    Let me know if you have any questions!
20. 20.▲
    by SilenN 2mo ago

    Thanks :)
21. 21.▲
    by SilenN 2mo ago

    Thanks for the heads up, removed mention!
22. 22.▲
    by SilenN 2mo ago

    Happy to answer any qs.
23. 23.▲
    by SilenN 2mo ago

    Valid criticism. Happy to answer any qs. We're still working on solidfying results.
24. 24.▲
    by SilenN 2mo ago

    It's open source! We do have a platform we'll be launching as well to manage training + serving for you which will require more diligent privacy guarantees.
25. 25.▲
    by SilenN 2mo ago

    Open source models. wmo routes requests between frontier models and open source models that continuously train using Tinker. As the smaller models improve, more traffic gets routed to them. Calculating cost is just tokens in/out.
26. 26.▲
    by SilenN 2mo ago

    That's cool, thanks for sharing!
27. 27.▲
    Show HN: Optimize and serve models with Fable quality at half the cost(github.com)

    71 points by SilenN 2mo ago|30 comments
28. 28.▲
    LLMs as 5x Faster Sandboxes(github.com)

    2 points by SilenN 3mo ago|1 comments
29. 29.▲
    by SilenN 3mo ago

    world-model-harness makes it easy to go from agent traces to faithful replication of your production environment where your agents run. Basically, an LLM pretends to be a virtual machine executing instructions. Based on GEPA and Qwen AgentW
30. 30.▲
    by SilenN 7mo ago

    Thanks! I do have a section on this in the article "Why genetic algorithms aren't state of the art" "Physics simulation involves discontinuities (contacts, friction regimes), long rollouts, and chaotic dynamics where sma