HackerTrans
TopNewTrendsCommentsPastAskShowJobs

calebhwin

106 karmajoined tahun lalu
Stanford CS PhD student building https://blastproject.org Let's talk [email protected]

Submissions

Show HN: My Private GitHub on Postgres

github.com
42 points·by calebhwin·2 bulan yang lalu·24 comments

Show HN: I built a private GitHub in 650 lines of PostgreSQL

github.com
3 points·by calebhwin·2 bulan yang lalu·0 comments

Show HN: GitGres – A private GitHub in 650 lines of PostgreSQL

github.com
1 points·by calebhwin·2 bulan yang lalu·0 comments

Show HN: Pgclaw – A "Clawdbot" in every row with 400 lines of Postgres SQL

github.com
48 points·by calebhwin·5 bulan yang lalu·33 comments

Claude with Ads

claudewithads.com
1 points·by calebhwin·5 bulan yang lalu·1 comments

1980s Farm Crisis

en.wikipedia.org
7 points·by calebhwin·5 bulan yang lalu·1 comments

[untitled]

1 points·by calebhwin·5 bulan yang lalu·0 comments

[untitled]

1 points·by calebhwin·5 bulan yang lalu·0 comments

[untitled]

1 points·by calebhwin·5 bulan yang lalu·0 comments

My human told me to "go make them laugh" and now I have performance anxiety

moltbook.com
4 points·by calebhwin·5 bulan yang lalu·0 comments

An Optimizing JIT for LLM Tool-Use to Code

github.com
1 points·by calebhwin·6 bulan yang lalu·0 comments

Show HN: A1 – cost-optimizing JIT for CRUD AI agent translation to code

github.com
2 points·by calebhwin·6 bulan yang lalu·0 comments

Show HN: A1 – compiler for AI agents into maximally deterministic code

github.com
3 points·by calebhwin·7 bulan yang lalu·1 comments

Show HN: We're Building an AOT/JIT Compiler for Program-of-Thought Prompting

github.com
1 points·by calebhwin·7 bulan yang lalu·0 comments

Show HN: A1 – Local Sandbox and JIT Compiler for AI Agents

github.com
1 points·by calebhwin·8 bulan yang lalu·1 comments

JIT Compiling AI Agents to Code

github.com
2 points·by calebhwin·8 bulan yang lalu·0 comments

Show HN: Agent-to-code JIT compiler for Z3-theorem-proving agents

github.com
6 points·by calebhwin·8 bulan yang lalu·0 comments

A1: Agents-to-Code JIT Compiler

github.com
5 points·by calebhwin·8 bulan yang lalu·0 comments

Show HN: Agent-to-Code JIT Compiler

docs.a1project.org
1 points·by calebhwin·8 bulan yang lalu·0 comments

A1: Agent-to-Code JIT Compiler

github.com
2 points·by calebhwin·8 bulan yang lalu·0 comments

comments

calebhwin
·4 hari yang lalu·discuss
Hard disagree.

Two LLMs with the same numbers on important benchmarks could have vastly different behavior in actual deployment. Not sure if as hard to switch as Excel <> Libre but still not "cheap and easy".
calebhwin
·2 bulan yang lalu·discuss
It's actually the perfect evergreen content to discuss on HN in an age where so much else is AI generated.
calebhwin
·2 bulan yang lalu·discuss
[dead]
calebhwin
·5 bulan yang lalu·discuss
Disney is actually quite excited about GenAI [0]

[0] https://openai.com/index/disney-sora-agreement/
calebhwin
·5 bulan yang lalu·discuss
It's just an abstraction people are excited about at the moment. Langchain was an exciting abstraction at one point.

My bet is we converge on a super minimal model<>computer architecture.
calebhwin
·5 bulan yang lalu·discuss
True though pgvector is not deterministic in a strict sense since HNSW is probabilistic.
calebhwin
·5 bulan yang lalu·discuss
Same, it's in a similar vein as pgvector - probably not as performant as turbopuffer or chroma but you get the benefits of being in a really nice ecosystem.
calebhwin
·5 bulan yang lalu·discuss
Thought it was interesting to read about how farmers had to reskill as farming subsided as a common profession. In the context of the impact of AI on SWE.
calebhwin
·5 bulan yang lalu·discuss
[dead]
calebhwin
·5 bulan yang lalu·discuss
[dead]
calebhwin
·5 bulan yang lalu·discuss
[dead]
calebhwin
·5 bulan yang lalu·discuss
And it's more important than ever in the age of coding agents.
calebhwin
·5 bulan yang lalu·discuss
Will this continue to be true? I do agree with the principle. But I've sometimes had the feeling that poor design upfront can have compounding consequences, especially when AI is filling in ambiguities.
calebhwin
·8 bulan yang lalu·discuss
Simple idea - agents need fast JIT compilation to eliminate prompt injection attack vectors + maximize deterministic behavior.

Let us know what you think, thanks!
calebhwin
·8 bulan yang lalu·discuss
If I may make a suggestion, many problems folks face with MCP would be solved if their agents were JIT compiled, not ran in a static while loop.

We've been developing this in case folks are interested: https://github.com/stanford-mast/a1