dbreunig

6 ms·

1.
My CCPA Dialog with OpenAI (2023)(dbreunig.com)
1 points by dbreunig 7d ago|0 comments
2.
What We Can Learn from Claude's Fable 5.1 System Prompt Changes(dbreunig.com)
4 points by dbreunig 10d ago|0 comments
3.
by dbreunig 25d ago
I think it’s a fine response when you say, “Doesn't feel well informed enough to give advice,” because I said 5.2
4.
Fable and the end of the free lunch(dbreunig.com)
225 points by dbreunig 25d ago|263 comments
5.
Harnesses Are "Situated Agents"(dbreunig.com)
1 points by dbreunig 1mo ago|0 comments
6.
AI to Kids by Letting Them Tweak Local Chatbots [video](youtube.com)
5 points by dbreunig 1mo ago|0 comments
7.
Why Open Source Matters for AI: Models should be infrastructure, not appliances(oreillyradar.substack.com)
3 points by dbreunig 1mo ago|0 comments
8.
by dbreunig 1mo ago
Wrote about this awhile ago, and it hasn’t changed: > Time and time again, when talking to people who rely on ChatGPT, Claude, Perplexity, and other general AI tools, I hear them say, “AI is incredible. It handles nearly everything I thr
9.
by dbreunig 1mo ago
> What you get is a beautiful animation that is 100% accurate and free of hallucinations How does he know?
10.
Dspy.Flex lets optimizers rewrite the code for faster, cheaper, better programs(cmpnd.ai)
1 points by dbreunig 1mo ago|0 comments
11.
Manage Your Agent's Loadout with Dr Skill(dbreunig.com)
2 points by dbreunig 2mo ago|0 comments
12.
Show HN: Brew doctor` for Skills and MCP loadouts(github.com)
2 points by dbreunig 2mo ago|0 comments
13.
by dbreunig 2mo ago
I got an early tip on this and used it to build a brand mood board and it crushed previous attempts with Claude. Highly recommend.
14.
Understanding the Dynamics of the AI Ecosystem with Pace Layers(dbreunig.com)
3 points by dbreunig 3mo ago|1 comments
15.
The Problem is Prompt Debt: You can't be model agnostic and hand-tune prompts(dbreunig.com)
4 points by dbreunig 3mo ago|0 comments
16.
When an agent can explain anything, what is the role of human-centric docs?(dbreunig.com)
4 points by dbreunig 4mo ago|0 comments
17.
by dbreunig 4mo ago
Elizabeth Lopatto at The Verge makes a strong case we _do_ have proof that Musk is actively gathering and throwing fuel on the fire: https://www.theverge.com/ai-artificial-intelligence/929129/s... > But the thi
18.
Overfitting to First Party Harnesses(dbreunig.com)
1 points by dbreunig 4mo ago|0 comments
19.
by dbreunig 5mo ago
Among benchmarkers its a frequent topic. Qwen BURNS reasoning to get its scores.
20.
Cybersecurity looks like proof of work now(dbreunig.com)
562 points by dbreunig 5mo ago|214 comments
21.
Claude Code builds a system prompt(dbreunig.com)
4 points by dbreunig 6mo ago|0 comments
22.
The Cathedral, the Bazaar, and the Winchester Mystery House(oreilly.com)
2 points by dbreunig 6mo ago|0 comments
23.
The 2nd phase of OSS in the agentic era: From clones to reimaginings(dbreunig.com)
2 points by dbreunig 6mo ago|0 comments
24.
The Cathedral, the Bazaar, and the Winchester Mystery House(dbreunig.com)
190 points by dbreunig 6mo ago|66 comments
25.
by dbreunig 6mo ago
Model testing and swapping is one of the surprises people really appreciate DSPy for. You're right: prompts are overfit to models. You can't just change the provider or target and know that you're giving it a fair shake. But
26.
Build a deep researcher and learn DSPy Signatures and Modules(cmpnd.ai)
2 points by dbreunig 6mo ago|0 comments
27.
Can chat bots accommodate advertising?(dbreunig.com)
2 points by dbreunig 7mo ago|0 comments
28.
by dbreunig 7mo ago
No reason it can't. I know people currently generating specs from existing code; just gotta write the pipeline.
29.
Learnings from a No-Code Lib: Keep the Spec Driven Development Triangle in Sync(dbreunig.com)
4 points by dbreunig 7mo ago|3 comments
30.
by dbreunig 7mo ago
"Think step by step," was just a sentence you appended to your prompt. It ended up kicking off reasoning training which enabled the massive gains in coding, tool use, and more over the last 18 months. So yeah, it's "just

More ›