MutedEstate45

6 ms·

1.
by MutedEstate45 1y ago
Good to know! Never been a maintainer before so I thought that was required.
2.
by MutedEstate45 1y ago
Agree with your pain points. One thing id add is GitHub makes you reapprove every PR after each push. As an OSS contributor it’s exhausting to chase re-approvals for minor tweaks.
3.
by MutedEstate45 1y ago
Great work! Definitely feels a lot more technical than the typical internship project.
4.
by MutedEstate45 1y ago
I'll look out for future iterations. Thanks and good luck with the paper.
5.
by MutedEstate45 1y ago
The headline feature isn’t the 25 MB footprint alone. It’s that KittenTTS is Apache-2.0. That combo means you can embed a fully offline voice in Pi Zero-class hardware or even battery-powered toys without worrying about GPUs, cloud calls, o
6.
by MutedEstate45 1y ago
I think you're conflating real-time monitoring with data retention. Zero retention means OpenAI doesn't store user data, but they can absolutely still filter content, rate limit and block harmful prompts in real-time without retai
7.
by MutedEstate45 1y ago
True, but there's still a meaningful difference in friction and scale. With closed APIs, OpenAI can monitor for misuse, throttle abuse and deploy countermeasures in real-time. With open weights, a single prompt jailbreak or exploit spr
8.
by MutedEstate45 1y ago
The repeated safety testing delays might not be purely about technical risks like misuse or jailbreaks. Releasing open weights means relinquishing the control OpenAI has had since GPT-3. No rate limits, no enforceable RLHF guardrails, no au
9.
by MutedEstate45 1y ago
Ah gotcha. Makes sense to get the benefits of modular adoption without the headaches. Nice approach.
10.
by MutedEstate45 1y ago
The modular ERP/MES/QMS approach is interesting and challenges traditional manufacturing processes. Most manufacturers obsess over single source of truth. (I.e. ensuring a part number means exactly the same thing across planning,
11.
by MutedEstate45 1y ago
Thanks. That Pareto trade-off is exactly what I'm trying to quantify not just qualify. For example, if I've got a $50 budget, what's the sweet spot? Scenario A: 100 agents × GPT-4o-mini × 500 steps Scenario B: 500 agents × lo
12.
by MutedEstate45 1y ago
Thanks for the clarification. That's a clean approach. I just stared your repo. Looking forward to seeing where sidequest.js goes :)
13.
by MutedEstate45 1y ago
I really like seeing the segmented buffer approach. It's basically the rope data structure trick I used to hand-roll in userland with libraries like fast-json-stringify, now native and way cleaner. Have you run into the bailout conditi
14.
by MutedEstate45 1y ago
Really like your approach of using existing Postgres/MySQL instead of dragging in Redis. It feels genuinely drop-in, but still Sidekiq-class. I know it's a bit early to ask about production patterns, but I was curious: if the work
15.
by MutedEstate45 1y ago
The real win isn't static vs dynamic typing. It's immediate, structured feedback for LLM iteration. cargo check gives the LLM a perfectly formatted error it can fix in the next iteration. Python's runtime errors are often con
16.
by MutedEstate45 1y ago
Interesting approach, but I'm curious about the practical cost considerations. A 1,000-agent simulation could easily be hundreds of thousands of API calls. The repo recommends gpt-4o-mini over gpt-4 and supports local Llama models, but
17.
by MutedEstate45 1y ago
I've built a couple of MCPs and what jumps out about this repo is the clean split into three MCP tools—unit, fuzz, and coverage—so an LLM can sequence tests just like any other commands. By keeping the AI layer thin and the orchestrati
18.
by MutedEstate45 1y ago
Thanks for the mention! I'm the author of Mcp-error-formatter, which was mentioned above. Exactly right about the client/server complexity. I've been working on this exact problem. Client-side heuristics can catch timeouts an
19.
by MutedEstate45 1y ago
Hi thanks for the suggestion! I think thats a great idea! I'll look into how to implement this
20.
Show HN: Mcp-error-formatter – Cursor-style JSON errors for any MCP/LLM tool(github.com)
4 points by MutedEstate45 1y ago|2 comments
21.
Show HN: System-info-now – Aggregate system debug data for LLM troubleshooting(github.com)
2 points by MutedEstate45 2y ago|0 comments
22.
by MutedEstate45 2y ago
Yeah ok that sounds really useful! Especially the sending test commands to other computers.
23.
by MutedEstate45 2y ago
Check out void, it's an open source alternative to Cursor and in early beta: https://github.com/voideditor/void/releases/tag/v1.0.0
24.
by MutedEstate45 2y ago
Ohhh thats encouraging! Was it also for mock data generation or did you make it handle other annoying tasks?
25.
Ask HN: A tool to handle annoying testing related tasks?
3 points by MutedEstate45 2y ago|4 comments
26.
by MutedEstate45 2y ago
I attempted the approach of having an AI write the code while I casually audited it, but the results didn’t work out well for me. It kept adding bugs and logical errors that were hard to find. I was using Cursor along with Claude 3.5 Sonnet
27.
by MutedEstate45 2y ago
Great work! The visual elements are very impressive. Would you say this tool is suitable for prototyping, especially for someone with development experience but no prior AR experience?
28.
by MutedEstate45 2y ago
Thanks for your suggestions, especially RocksDB! I’m gonna play around with it today. Do you have any ideas related to use cases? I was thinking something like dynamic RAG for transformers.js
29.
Ask HN: Why Try RAG on the Client Side? Is Local-First RAG Practical?
4 points by MutedEstate45 2y ago|2 comments
30.
HumanRankEval: Automatically Evaluating Conversational AI Output(aclanthology.org)
2 points by MutedEstate45 2y ago|0 comments

More ›