freelance automations
Mar 28 · 12 min read · Your AI agent has been running for 47 days straight. Zero crashes. Zero timeouts. The uptime dashboard shows 99.97%. Your SRE team is satisfied. Meanwhile, the agent has been hallucinating tool parameters for the last six hours. It confidently genera...
Join discussionMar 28 · 8 min read · Your AI agent has a temperature setting in a YAML file, a model name in an environment variable, a system prompt in a Python string, tool definitions in a JSON schema, retry limits in a constants module, and feature flags in a database. You change th...
Join discussionMar 26 · 12 min read · Deploying a new version of your AI agent is not the same as deploying a new version of a web service. You can roll back bad code. You cannot easily roll back bad behavior that corrupted 3,000 records or sent 500 emails before anyone noticed the model...
MEAamer and 1 more commentedMar 25 · 9 min read · Every time your agent processes a user query, it calls the LLM. Every time it encounters a similar query, it calls the LLM again. Same system prompt, nearly identical user message, same model, same result — but a fresh API call every time. At $15 per...
EAli commentedMar 22 · 11 min read · Most agent code I review looks like this: the LLM client is instantiated inline, the vector store connection is hardcoded in the tool function, the prompt template is a raw string in the agent loop, and the embedding model is imported at the top of t...
Join discussion