BMD HODL devlog - week of 2026-05-17
From 2026-05-10 through 2026-05-17 I tightened the AgentGuard story, cleaned up site trust issues, and got a cleaner read on where autotrader still lags passive.
BMD HODL devlog - week of 2026-05-17
From 2026-05-10 through 2026-05-17, the biggest move was getting stricter about proof. I cleaned up the AgentGuard funnel and numbers on bmdpat, kept agent47 in maintenance mode, and let autotrader tell the truth instead of forcing a hero story. The headline is simple: I shipped useful surface area, but the more important win was making the stack harder to fake. AgentGuard still has no real external pull yet. Autotrader still trails passive. That is useful because it keeps me pointed at the real problems.
What shipped
bmdpat
- PR #398: fixed the Bazaar x402 extension shape so indexed services see the right contract.
- PR #418: replaced the inflated AgentGuard download fallback with live pypistats.
- PR #419: rewrote the AgentGuard landing page to lead with MCP-native governance.
- PR #420: narrowed the lifetime AgentGuard counter to Pepy-only instead of mixing incompatible totals.
- PR #421: fixed blog excerpt metadata.
agent47
- PR #467: improved the SDK first-run proof so the product story starts with a cleaner success path.
- PR #472: refreshed MCP indexing state and documented the blockers instead of pretending registry drift was fixed.
autotrader
- PR #16: added the CBRS post-IPO watchlist update for the paper-only V2 book.
- PR #17: logged the CBRS watchlist and power or colo queue note into the inbox flow.
- PR #18: landed the power and colo sentiment watchlist for GEV, VST, CEG, and ANET.
- PR #20: landed the stranded 2026-05-13 regime and watchlist knowledge updates.
What I learned
stratechery-inference-shiftplustomtunguz-localmaxxing: local inference economics are now part of the product thesis, not side research.claude-code-programmatic-restrictions-2026-05-14: if an agent workflow depends on bundled pricing, I need the meter-read before I trust the lane.opensquilla-token-cost-agent: runtime spend control is getting crowded, so AgentGuard needs proof of enforcement, not just download vanity.
Numbers
- Autotrader: combined book +6.4% as of 2026-05-16, with -5.1% alpha vs SPY on stocks and -10.3% alpha vs BTC on crypto.
- Closed loops: 0 install intents and 0 CTA clicks in the 7-day readout dated 2026-05-17.
- AgentGuard PyPI: 483 installs over 7 days in the 2026-05-17 focus review. The mirror-aware readout still disagrees, so the metric needs one scraper audit.
If you want hard budget limits and loop guards for coding agents, start here: https://bmdpat.com/tools/agentguard
Get the local AI lab notes
Benchmark rows, VRAM fit checks, quant choices, and what actually runs on consumer GPUs. M-F, only when there is something worth sending.
Patrick Hughes
Building BMD HODL — a one-person AI-operated holding company. Nashville, Tennessee. Twenty-Two agents.
More writing
- 8 min
BMD HODL devlog - week of 2026-05-03
I spent the week tightening the AgentGuard release path, shipping proof-heavy docs and perf fixes, and keeping the benchmark gap visible instead of hand-waving it away.
- 11 min
BMD HODL devlog - week of 2026-04-26
Closed loops shipped, the funnel got narrower, and the weekly scoreboard got more honest.
- 4 min
When Your Blog Repair Loop Fails 23 Times, Stop Repairing
My blog repair loop chewed on a stale draft for 23 mornings and reported "blocked" every time. The fix was not a smarter retry. It was a TTL and a heal path.