Stories by dbreunig

Astra can read and write weights to improve Qwen

Show HN: A working Nintendo in ~1GB of Qwen weights

The State of Context Engineering

Context Inertia: When Agents Won't Let Go

ReAnchor, a Jev Optimizer in DSPy

The Rise of Audio AR (2024)

My CCPA Dialog with OpenAI (2023)

What We Can Learn from Claude's Fable 5.1 System Prompt Changes

Fable and the end of the free lunch

Harnesses Are "Situated Agents"

AI to Kids by Letting Them Tweak Local Chatbots [video]

Why Open Source Matters for AI: Models should be infrastructure, not appliances

Dspy.Flex lets optimizers rewrite the code for faster, cheaper, better programs

Manage Your Agent's Loadout with Dr Skill

Show HN: Brew doctor` for Skills and MCP loadouts

Understanding the Dynamics of the AI Ecosystem with Pace Layers

The Problem is Prompt Debt: You can't be model agnostic and hand-tune prompts

When an agent can explain anything, what is the role of human-centric docs?

Overfitting to First Party Harnesses

Cybersecurity looks like proof of work now

1 2 3 4 5 6 7 8 9 10