21 posts
AI-Driven Development isn't a replacement lifecycle — it re-wires the one you already run. A stage-by-stage field guide mapping ADD onto requirements, design, dev, review, test, security, release, ops, and maintenance — grounded in two real projects.
19 min readA data-driven companion to the ai-proxy field study. Mining the transcripts and the .add foundation to measure an AI-driven build across the full SDLC and every role, against modeled human baselines of 6 person-months to 32 person-years, with the caveats that keep the comparison honest.
27 min readAI-Driven Development was born in software, but the method is domain-independent. The producer is now an AI in every knowledge-work domain — so the same loop applies: constrain the what, free the how, verify by evidence. The framework for the whole series.
11 min readAI support drafts are fluent and frequently wrong — invented policy, steps that don't exist. ADD makes the source-of-truth the contract, factuality-against-source the red test, and a measured hallucination rate the evidence — so confident wrong answers are caught before they reach a customer.
10 min readAI can emit Terraform and runbooks that read correct and provision the wrong, insecure, or irreversible thing. ADD makes policy-as-code the frozen contract enforced in the pipeline, chaos and rollback tests the red tests, and SLO evidence — not a clean apply — the proof a change is safe.
11 min readCompliance is where ADD's gates fit most naturally. AI control documentation reads thorough and can map to nothing real; ADD names every violation, freezes the control framework, makes control tests the red tests, and treats every finding as a hard stop.
13 min readWhen agents and teams produce more than any leader can read, line-by-line review collapses. ADD reframes the job: the ADR and its fitness functions are the frozen contract, fitness checks are the red tests, and production evidence — not a clean diagram — is how direction is verified.
13 min readAn AI-built model reconciles at a glance and hides a broken assumption or a double-count. ADD locks the assumption set as the frozen contract, makes reconciliation and integrity checks the red tests, and verifies by backtest against actuals — not by whether the numbers look reasonable.
15 min readAI screening reads objective and can be quietly biased. ADD makes the calibrated rubric the frozen contract, a gold-set plus counterfactual bias probes the red tests, and adverse-impact tracking the evidence — so a human, never the model alone, owns every decision about a person.
11 min readML is AI building AI, and a high headline metric hides leakage, gamed proxies, and slice regressions. ADD locks the eval set, thresholds, and model card as the frozen contract, makes the held-out suite the red tests, and verifies by slices and online evidence — not by the leaderboard number.
14 min readLegal is where ADD maps most literally — the negotiation playbook is the frozen contract. AI drafts and redlines read authoritative and can quietly concede a material term; ADD makes playbook-conformance the red test and an adversarial redline the evidence, with off-playbook concessions a hard stop.
12 min readAI floods marketing with on-brand-looking copy that is off-strategy. ADD fixes it: the brief is the spec (with named refusal reasons), brand and claims are the frozen contract, the A/B test is verification by evidence, and performance folds into a living playbook.
11 min readAI can generate plans and status that look complete while the goal is unmet. ADD makes evidence-backed exit criteria the contract, the dynamic goal-loop the engine, and an artifact — not a self-reported 'green' — the proof a milestone is actually done.
12 min readAI can flood prospects with personalized-looking outreach that is off-ICP, mis-priced, or non-compliant. ADD clamps the what — ICP, claims, pricing authority — and frees the how, then verifies by reply and win rates instead of by how persuasive the copy reads.
13 min readA field study of ai-proxy, a multi-tenant AI gateway built end-to-end with ADD: 23 milestones, ~120 tasks, six days, zero waivers. What the method felt like in production and how the LLM behaved when scope was clamped.
12 min readADD's first two steps kill fast waste before a line of code exists: Specify (Must do, must Reject with error codes, the After-state) and Scenarios (concrete happy, edge, and failure examples in domain language). Pin the what.
11 min readTests-first, pointed at an AI agent: write a red suite that asserts observable behavior, then turn the agent loose with one instruction — make every test pass, change nothing in the tests or the contract. Red to green.
11 min readAI made writing code nearly free and quietly broke the SDLC. AI-Driven Development — five competencies, eight steps — constrains what an agent builds while freeing how it builds, proven across a 23-milestone production gateway shipped in six days.
26 min readBefore RAG, before tools, before agents — a well-written system prompt is still the single highest-leverage knob you have. Here's what it actually does, why most people misuse it, and what its honest limits are.
9 min readEvery AI product in the world starts as the same small, strange thing — a stateless function that turns tokens into tokens. Understand that substrate clearly and every capability above it stops looking like magic.
9 min readA chatbot answers. An agent finishes the job. The gap between them is not one feature — it's a stack of seven capabilities, built one rung at a time. This series walks that stack in plain English, for engineers and curious non-engineers alike.
9 min read