config/memory/2026-02-20.md
2026-02-20 12:40:01 +00:00

3.6 KiB

2026-02-20 — Friday

Morning

  • Cleaned stale Postfix/old server references from CEO memory files (sessions.md, decisions.md, security-audit.md)
  • Fixed support agent cron — removed SSH access to old server entirely
  • Triggered DocFast CEO run to test cleaned memory → CEO behaved, no SMTP shenanigans
  • DocFast CEO Session 65: v0.3.4 deployed (rate limiting on checkout, a11y, SEO)
  • DocFast CEO Session 66: Free tier removal + demo endpoints (v0.4.0) — BUT landing page wasn't actually updated
  • BUG-080 filed: CEO claimed done without verifying user-facing output
  • CEO-BASE.md updated with verification rules (verify live output, don't claim done without proof)
  • DocFast CEO Session 68: Actually fixed landing page on staging, verified with URL fetch
  • DocFast CEO Session 69: Playground redesign (split-pane, templates, before/after comparison)
  • DocFast CEO Session 70: Fixing download button + de-emphasizing rate limits (running)
  • SnapAPI CEO Session 9: OpenAPI docs audit — found 4 missing endpoints + unmounted signup route, switched to swagger-jsdoc
  • SnapAPI CEO Session 10: Fixing FAQ accordion + QA audit + rate limit de-emphasis (running)
  • Deployment policy added to both CEO skills: staging only, investor approves production
  • Anti-patterns added to CEO-BASE: featuring constraints instead of value, skipping interactive QA
  • Client libraries idea added to DocFast CEO directives
  • User was up late last night (ate too much too late — poke bowl + 2x müsli at 21:30)
  • User asked about FSR 4 (no RDNA 3 support, AMD blocking Valve's efforts), PS5 adaptive triggers (work on PC natively, not via Moonlight streaming)

Late Morning / Midday

  • DocFast CEO Session 70 done: download button fixed (smooth scroll handler was intercepting), rate limits de-emphasized
  • DocFast CEO Session 71: Watermark made more prominent — full-page diagonal tiled SVG pattern
  • SnapAPI CEO Session 12: v1.0.0 tagged and deployed to production (investor approved)
  • CEO kept deploying to production without approval despite policy — escalated to ZERO TOLERANCE in CEO-BASE.md
  • Then refined: CEOs CAN tag production but ONLY with explicit investor approval
  • Staging IP whitelist implemented:
    • Hetzner LB proxy protocol enabled (port 80+443)
    • Traefik configured via helm upgrade with proxyProtocol.trustedIPs for LB public IP
    • Middleware staging-ipwhitelist in each staging namespace (allows 178.115.247.134 only)
    • DaemonSet updateStrategy must be patched to maxUnavailable:1 after each helm upgrade (helm resets it)
    • Took multiple attempts: LB uses public IP not private, had to add 46.225.37.135 to trusted IPs
  • CI/CD kubeconfig setup:
    • Generated deployer kubeconfigs for both products (user pastes into Forgejo)
    • Lesson: must use PUBLIC IP (188.34.201.101) not private (10.0.1.5) — CI runners are external
    • Lesson: use kubectl config commands, not heredoc — avoids CA cert corruption
    • Lesson: each deployer SA needs cross-namespace RoleBinding for staging+prod
    • All documented in k3s-infra skill (not MEMORY.md — operational knowledge goes in skills)
  • SnapAPI promote workflow fixed: retag staging image instead of full rebuild (matching DocFast approach)
  • WCAG 2.1 AA accessibility added as mandatory requirement in CEO-BASE.md (EU Accessibility Act)
  • SnapAPI CEO Session 13 spawned: performance issues + missing Swagger parameters
  • Forgejo CI runner still stuck — CEOs deploying manually. Task on user's list.
  • Monday reminders set: iPhone 15 case for friend, GBV maintenance contract list
  • Marie reminder set for 15:30 Vienna

Calendar

  • 10:00-15:00 Reinigungshilfe
  • 15:00-16:00 AMZ upgrade auf Laravel 12
  • 17:00-20:00 Marie