m@berryhill: ~/berryhill.dev/posts
~ home posts/ about.md
guest@berryhill : ~ $cd posts/ && ls -la --sort=date

posts/archive

Long-form notes on agentic-first development, AI / ML systems, blockchain rails, crypto market structure, digital music, and the small frictions that change when intelligent automation moves into the loop.

total113 posts
words244,438
topics7
first post2025-10-27
last post2026-09-24
cadencelive
~/posts $ ↵ to apply · esc to clear
all 113 ai-agents 21 evaluation 10 governance 3 agent-ops 6 parallel-test-time-scaling 1 long-horizon-agents 1 compaction 1

cat pinned.md

2 starred · author's pick
.md ★ 12 min · 2026-07-07
the-agent-stack-is-getting-protocols-before-it-gets-governance.md
MCP, A2A, and agent SDKs make agents easier to connect. Production trust depends on the control layer above them: authority, review, evidence, recovery.
topic AI Agents words 2,704 open →
.md ★ 14 min · 2026-07-03
the-agent-stack-governance-checklist-authority-review-provenance-and-handoffs.md
A practical checklist for designing authority, review, provenance, and handoff gates around MCP, A2A, and agent SDK workflows before they touch production work.
topic AI Agents words 3,147 open →

ls -la 2026/

6 posts · sorted by timestamp
$ cd 2026/ 12,999 words
2026-09-24 7 min population-monitoring-as-an-agent-warning-signal.md Individual agent checks can miss collective drift. Here's how to evaluate a population-level warning signal without mistaking it for a safety guarantee. → 2026-09-23 7 min compaction-flips-the-model-tier-tradeoff.md A v1 preprint on cost-efficient compaction for long-horizon coding agents reports >10pp Terminal-Bench uplift for less than the cost of two full-context runs. Pricing compaction at the rollout boundary, not end-to-end, lets a cheaper model match a flagship. → 2026-09-22 11 min what-can-you-remove-after-harness-distillation.md Harness distillation can make some runtime guidance removable. Use a retain-or-retire map to separate learned habits, tool interfaces, and enforced controls. → 2026-09-20 14 min scoring-definition-is-the-lever-in-agent-evals.md A 24.0%-57.0% consensus swing on identical held-out data is a property of the scoring definition, not the agents. Lock the definition before reading the number. → 2026-09-19 11 min spreadsheet-rag-can-find-the-right-wrong-chunk.md A spreadsheet retrieval hit can lose the headers that give a value meaning. Separate missing evidence, missing context, and incorrect answers. → 2026-09-18 10 min competent-agents-can-still-fail-to-coordinate.md Agent capability is not proof of a valid shared commitment. What scheduling experiments suggest builders should test across different owners. →
$ grep returned no matches. try a different query or reset
page 1 · showing 1-6 of 113 live posts.

cat subscribe.md

stay in the loop
RSS= /rss.xml · standard feedsubscribe → CONTACT= selective consulting · no other reasonopen →

guest@berryhill : ~/posts $