Hermes + Ollama on a 16GB Mac: a 30-minute setup
Step-by-step from zero to a Telegram bot that answers questions using llama3.1:8b. With the gotchas around context size, tool streaming, and the launchd service.
Coming soon · 8 min read
Field notes
Tutorials, gotchas, and benchmarks. Stub posts for now — full writeups land here as I work through them. Subscribe via the form at the bottom to get new posts by email.
Step-by-step from zero to a Telegram bot that answers questions using llama3.1:8b. With the gotchas around context size, tool streaming, and the launchd service.
Coming soon · 8 min read
I ran both models against the same 400-line Flask PR. Same prompts, same OpenCode config. Surprising result on multi-file reasoning.
Coming soon · 12 min read
After 3 months of running both daily, here's the practical split. They have feature overlap, but the day-to-day UX differs enough that I keep both alive.
Coming soon · 6 min read
Function calling, structured output, vision, embeddings — the /v1 endpoint is "compatible" but there are sharp edges. Here's the compatibility matrix I wish someone had given me.
Coming soon · 10 min read
Stay in the loop
No newsletter platform, no tracking, no schedule promises. Just an email when a new post lands.