AWS Veteran: The New Software Development Life Cycle

<p>&quot;I need to stop using Opus. This doesn&#39;t work.&quot; That was Heitor Lessa&#39;s conclusion after a refactor cost him 200 million tokens, and it forced him to rebuild the entire agent workflow now available for 1400 engineers. Heitor spent 11 years at AWS, built Lambda Powertools to 230 billion API calls a week, and in this episode he walks through the full SDLC workflow on screen, from discovery to merge check.</p><p>In this episode, we cover:</p><ul><li>The product loop: discovery, whiteboarding, and the /roadmap command</li><li>Spec-driven development with Open Spec and why vanilla setups fail</li><li>Three model tiers: SOTA for planning, mid-tier for implementation, cheap models for reviews</li><li>Merge checks with adversarial reviewers and attestations that catch agents fabricating test results</li><li>The /retro command: using the Socratic method to make your workflow more deterministic</li></ul><p>If you&#39;re an engineer figuring out how to work with agents at team scale without losing trust in your codebase, this is the workflow to steal. This is also the first Beyond Coding episode with visuals on screen, so let me know what you think of the format.</p><p>Timestamps:<br>00:00:00 - The Math Doesn&#39;t Add Up<br>00:00:43 - Amazon Hypergrowth: 11 Years, 8 Different Roles<br>00:03:29 - Learning From the Trenches as a Technical Account Manager<br>00:08:38 - Developer Identity and the Birth of Lambda Powertools<br>00:10:20 - The Hard Parts of Working in Public<br>00:13:12 - How Powertools Hit 230 Billion API Calls a Week<br>00:16:42 - Career Advice: Learn Adjacent Roles, Not More Tech<br>00:19:37 - When Leadership Decisions Don&#39;t Make Sense to You<br>00:23:21 - The Product Loop Starts With Discovery<br>00:25:22 - From Whiteboard to /roadmap<br>00:27:37 - Why Humans Plan First and Agents Come Second<br>00:30:33 - Commands vs Skills Across 32 Different Models<br>00:33:38 - Adversarial Reviewers on Every Plan<br>00:36:07 - The Socratic Method, Explained<br>00:40:29 - Why He Only Takes Paper Notes<br>00:44:43 - The Five-Line Paper Trick for High-Stakes Meetings<br>00:48:18 - /new-work: Capturing Scope Creep Without Derailing<br>00:54:03 - The Dev Loop Begins: Open Spec Explore<br>00:56:34 - Three Model Tiers: SOTA, Mid, Cheap<br>00:57:43 - The $5,000/Month Per Engineer Question<br>00:58:57 - Guardrails vs Autonomy for 1,400 Engineers<br>01:04:22 - Auto-Sizer: Does This Task Even Need a Spec?<br>01:07:26 - Decision Fatigue and Why Frameworks Win<br>01:09:10 - The Plan Phase: Specs, Design, Formal Verification<br>01:13:07 - The Refactor That Cost 200 Million Tokens<br>01:15:11 - When Agents Forge Evidence They Ran Your Tests<br>01:17:27 - Local-First Architecture Explained<br>01:23:04 - The Apply Phase: Fully Autonomous Loops<br>01:24:30 - Coding Was Never the Bottleneck<br>01:26:39 - Why This Workflow Is an Investment<br>01:27:39 - Decision Logs and the /onboarding Command<br>01:29:06 - Running Agents Locally With Enterprise Governance<br>01:32:42 - Hooks: Making Quality Gates Deterministic<br>01:36:02 - Merge Checks: 15 Adversarial Reviewers Per Change<br>01:38:30 - /retro: Interviewing Yourself to Improve the Loop<br>01:43:12 - Trust, Loss of Trust, and Recovery With Agents<br>01:48:02 - Experience, Scars, and Critical Thinking<br>01:49:32 - Why Right Now Is the Time to Experiment<br>01:52:04 - Conviction Comes From Being in the Loop</p><p><a href="#softwareengineering" target="_blank" rel="ugc noopener noreferrer">#softwareengineering</a> <a href="#aiagents" target="_blank" rel="ugc noopener noreferrer">#aiagents</a> <a href="#aws" target="_blank" rel="ugc noopener noreferrer">#aws</a></p>