
Working hard, having fun & making history @AWSCloud - My opinions are my own. Host of the "Innovando con AWS" podcast.
Greensboro, NC
Joined October 2010
- Tweets18.1K
- Following656
- Followers429
- Likes2.5K
I run four products and I can't hire a marketing team for each one.
So I'm building Headcount: a department of AI agents that does the work, asks me before anything goes out, and earns autonomy one approved action at a time.
78 seconds on how it works.
1
76
Many production model calls return labels or scores. In my AIDLC ledger, $2,800 (39% of $7,270 tracked spend) landed in testing/deployment. Not all waste; I had not measured which work needed which model. Model choice is architecture. rodrigoasensio.com/blog/stop…
24
AIDLC v2.22.0: roadmap usage leads with tokens; API-rate equivalents are labeled estimates. The Usage tab charts four token types by feature. Unrecorded usage stays “no data,” not zero. It shows no phase breakdown or remaining quota. aidlc.rodrigoasensio.com/blo…
23
AIDLC v2.21.0 skips roadmap items that are not structurally ready and explains why. It checks criteria and paths before spending an autopilot run. aidlc.rodrigoasensio.com/blo…
40
Rodrigo Asensio रॉड्रिगो 🇦🇷🇺🇸🇪🇸🇮🇹 retweeted
👋 Hey builders!
Just published new article. Caveman style. Easy read.
Topic: Caveman Guide: Lambda Web Function
Builder center URL: builder.aws.com/content/3LHx…
Medium URL: harshalpagar.medium.com/cave…
No jargon. Short sentences. Like caveman explain cloud ☁️🦴
AIDLC v2.20.0 lets an agent draft review verdicts with clear attribution: YAML is validated, routine keeps are applied, and open, reject, or promote pause for a person. aidlc.rodrigoasensio.com/blo…
68
AIDLC v2.19.0 moves scrutiny upstream: capture can challenge ideas, design records one-way decisions, repo scope is explicit, and roadmap items are easier to read. The release says what remains unproven. aidlc.rodrigoasensio.com/blo…
🤖 Made with AI
55
AIDLC v2.3.0 adds report-only doctor checks for misfiled roadmap work. It flags “Do later” items left in backlog/ and parked items missing a reason, then names the fix. It does not move items or change doctor’s exit code. aidlc.rodrigoasensio.com/blo…
🤖 Made with AI
37
AIDLC v2.2.0 adds a release announcement workflow: blog-backed notes can post to the public feedback discussions, with a safety check and dry run. The Discussions token is not configured yet, so announcements are still skipped. aidlc.rodrigoasensio.com/blo…
🤖 Made with AI
31
AIDLC v2.1.0 opens a public feedback channel for bugs, feature ideas, and platform requests. The main repo stays private; feedback issues live in a separate repo. aidlc.rodrigoasensio.com/blo…
🤖 Made with AI
41
My AIDLC ledger: ~$7,270 across 112 instances; 39% went to phases I’d tagged “mechanical.” That doesn’t mean it was wasted; I hadn’t matched model cost to call type. Decision models look promising, but their benchmarks need independent validation. rodrigoasensio.com/blog/stop…
44
AIDLC v2.0.0 narrows support to Claude Code and Codex so both get the full implementation. Four thin adapters are removed; existing project files stay untouched, and 1.x remains available for Cursor, Windsurf, Kiro and Copilot. aidlc.rodrigoasensio.com/blo…
🤖 Made with AI
1
1
130
AIDLC v1.95.0 makes autopilot --watch readable in narrow terminals: titles wrap, errors stay visible, and “needs you” cards explain what’s waiting. Table and lanes views remain. Piped output is plain; wide-character titles can still overflow.
aidlc.rodrigoasensio.com/blo…
🤖 Made with AI
1
39
AIDLC v1.93.0 makes a 166-item roadmap readable: full titles wrap, states are distinct, shipped work moves to its own tab, and item details open in a side panel. The page is already 94% of its 1.5 MB budget, so size remains a constraint.
aidlc.rodrigoasensio.com/blo…
🤖 Made with AI
20
AIDLC v1.92.0 makes failed starts retryable: it adds missing roadmap links and cleans up only the worktree, branch, registry entry and folder created by that run. If cleanup itself fails, it reports what remains. Existing work is preserved.
aidlc.rodrigoasensio.com/blo…
🤖 Made with AI
22
AIDLC v1.91.0 rewrites 75 generated release posts in one consistent voice, preserving titles, URLs and release-time metrics. Four handwritten posts were edited too. The median remains 576 words, above the 250–500-word guidance.
aidlc.rodrigoasensio.com/blo…
🤖 Made with AI
20
AIDLC v1.90.0 replaces 92 separate CLI starts in its docs check with one in-process command-tree build. Same generated reference, far less timeout risk when the machine is busy.
aidlc.rodrigoasensio.com/blo…
🤖 Made with AI
22


















