LLM Software Development Pitfalls: What to Avoid and When to Simplify
6d ago · 19 min read · TLDR: Most bad LLM products do not fail because the model is weak. They fail because teams wrap a maybe-useful model in too much architecture: prompt spaghetti, no eval harness, weak tool schemas, huge context windows, agent chains nobody can explain...
Join discussion
Twenty-One Documents and a Weak Reject: Building a Research Corpus for a Novel Search Architecture\n
Apr 16 · 11 min read · Every vector search system you have ever used assumes the same thing: that someone already ran every document through an embedding model and stored the results. For most workloads, that assumption is fine. For some, it is fatal. Forensic seizure of 5...
Join discussion
Skaro 2.0: Not Another AI Coding Tool, but a Workspace for Building Software With AI
Apr 13 · 9 min read · Skaro 2.0: Not Another AI Coding Tool, but a Workspace for Building Software With AI I wrote about Skaro before, but since then the project has changed quite a lot. Version 2.0 is not a cosmetic updat
Join discussion
Stop Training Individuals. Start Building Context Infrastructure.
Apr 12 · 9 min read · This is the full breakdown behind my LinkedIn post on AI-native developer workflows. It's longer — deliberately. The short post tells you what I built. This one tells you why, how it works, and what i
Join discussion
When AI Does the Thinking: What Agentivism Teaches Us About Kids, Learning, and the Delegation Trap
Apr 11 · 19 min read · The Question Parents Are Getting Wrong The conversation about children and AI tends to split into two camps that talk past each other. One camp focuses on content safety — can the model produce inappr
Join discussion
