llama.cpp sessions with coding agents
11 sessions on llama.cpp with a coding agent: the agent failed at least once in 7, and 4 finished without a human stepping in.
Plan and scaffold a Gradio hackathon app for tuition teachers
Takeaway: I'd start next time by pasting in the full sponsor/bonus badge list alongside the PRD in the first prompt, since that reshaped the whole plan later and cost a lot of back-and-forth.
Implement Phase 6 badge features for TutorDesk AI (Gradio 6, llama.cpp, HF traces)
Takeaway: I'd start the next session by asking for the smoke test script up front, since testing all four models turned out to be a separate need I only realized after Phase 6 was 'done'.
Update docs to reflect curriculum-audio local TTS path
Takeaway: Next time I'd bundle the doc-consistency check into the original implementation prompt instead of asking for it separately, since the docs drift as soon as behavior changes.
Get Off the Grid local inference badge-ready on a HF Space
Takeaway: Next time I'd lead with the Space's hard constraints (1GB repo limit, free-tier build memory, no CUDA) before letting the agent draft a plan, since two of our failures were pure packaging/build issues we could have anticipated upfront.
Audit, clean up, and deploy a Read-Along AI Gradio app to a Hugging Face Space
Takeaway: Next time I'd ask upfront for a deploy script instead of letting the agent hand-run git/HF CLI commands each time — that's what actually stabilized the workflow.
Add local GGUF inference path and hybrid Modal/local toggle for read-along app
Takeaway: Next time I'd ask for the environment sanitization (Conda vs .venv Torch conflicts) to be handled up front, since that same libtorch_python.so issue bit me twice, once during GGUF conversion and again during local_inference.py testing.
Rework project embeddings to use Gradio app_file text and remove 100-project limit
Takeaway: Next time I'd open with both asks together — what goes into the embedding and why there's a 100-project cap — since the agent had to dig through the same files twice.
Build 6-agent Kirana Detective invoice auditor with YOLO26n + MiniCPM fine-tuning
Takeaway: Next time I'd flag the Windows-specific gotchas (WeasyPrint's native deps, SQLite file locks, libGL missing on Modal's slim image) up front in the first prompt since they cost real cycles to debug reactively.
Debug Modal fine-tuning script for MiniCPM5-1B and migrate HF repo paths
Takeaway: Next time I'd paste the full Unsloth/Modal error and ask for the actual root cause up front instead of letting quick patches loop on the same symptom.
Push fine-tuned MiniCPM-V invoice model to HuggingFace and fix inference format
Takeaway: Next time I'd start by checking the app's actual inference stack (llama-cpp-python, not vLLM) before pushing any model format, since that assumption cost a lot of back-and-forth on GGUF conversion that turned out to be a dead end for MiniCPM-V-4.6.