ed@edheltzel:~/thelog$ ls -ahl
The Log
2026
- Sep 12, 2026The actual pain points of agentic engineeringAfter six months of daily agent use, here's where it breaks down
- Sep 10, 2026Clankers in your codebaseWhat happens when you wire a token launcher into your deploy pipeline
- Sep 8, 2026Fine-tuning MedGemma on your own dataPractical notes on adapting Google's medical vision model with LoRA
2025
- Aug 29, 2025Error recovery needs a stopping ruleFeeding a stack trace back to the model can unstick a call. Doing it forever is how agents spin.
- Aug 25, 2025Human approval belongs before the side effectIf the model can select a dangerous tool, you need a break in the loop before that tool runs. Slack is a channel, not an afterthought.
- Aug 21, 2025An agent should be able to stop and come backIf you cannot serialize the thread, you do not have an agent you can pause. You have a process you hope stays up.
- Aug 18, 2025The context window is an application interfacePrompts and context are code. If a framework hides the tokens, you cannot tune the only interface the model has.
- Aug 15, 2025Your agent does not need to own the workflowMost of a good agent is still your code. The model only names the next structured step.