# ng-thebatch-359-3-product-development-loops-2026-06-26

## Veille

Letter "Dear friends" from Andrew Ng in *The Batch* (DeepLearning.AI, issue 359) on **loop engineering** applied to **0-to-1** product development. Ng shares his **3 key loops** — agentic coding loop (~minutes), developer feedback loop (~hours), external feedback loop (~days) — nested by increasing time scale, connecting *coding agent → product spec/evals → developer vision → external feedback*. Central thesis: humans retain a **context advantage** (rather than a "taste") that makes human-in-the-loop indispensable; engineers take on a partial product management role. Domain: coding agents, product engineering, agentic methodology.

## Titre Article

3 Key Product Development Loops (The Batch, Issue 359 — « Dear friends » letter)

## Date

2026-06-26

## URL

https://www.deeplearning.ai/the-batch/issue-359

## Keywords

Loop engineering, product development, agentic coding loop, developer feedback loop, external feedback loop, evals, product spec, product vision, context advantage, human-in-the-loop, taste, product management, coding agents, 0-to-1

## Authors

Andrew Ng

## Ton

**Profile**: personal first-person letter ("Dear friends… Keep building! Andrew"), warm and pedagogical register, addressed to a community of builders. Accessible technical level, anchored on a single diagram (3 concentric loops, ~minutes / ~hours / ~days scales).

**Style**: didactic and encouraging, illustrated by a running **personal anecdote** (the keyboard-typing learning app built over the weekend for his daughter — unlockable cat costumes, parent login flow, visual design changed several times). Ng proceeds by framed definitions followed by lived examples, with a deliberate conceptual nuance: he **prefers "context advantage" over "taste"** because it "gives a clearer path to help AI systems improve." Authority: a reference figure in AI (founder of DeepLearning.AI, Coursera, Google Brain), speaking as a practitioner ("over the weekend, I was building…"). Target audience: engineers and builders adopting coding agents and growing into a product role. Forward-looking tone ("I will write more about how to do this in future letters").

## Pense-betes

- "**Loop engineering**" became a viral *buzzphrase* after mentions by **Boris Cherny** (creator of Claude Code) and **Peter Steinberger** (creator of OpenClaw); loops are now key to keeping agents iterating on software for a long time.
- Ng describes **3 nested loops** that guide *how* he builds **and** *what* to build. Chained nodes: `coding agent → product spec/evals → developer vision → external feedback`.
- **Loop 1 — agentic coding (~minutes)**: `coding agent ↔ product spec/evals`. Given a spec (+ optional evals), the agent writes code, tests it, iterates until it is bug-free and compliant with the spec. "Closing the loop" took off **late last year** → *game changer* (agents productive longer without intervention). Example: the agent worked **~1 hour**, using a browser to check its work several times, without intervention. Runs every few minutes. **Active area of invention.**
- **Loop 2 — developer feedback (~tens of minutes to hours)**: `product spec/evals ↔ developer vision`. The developer reviews the product and **steers** the agent. Last year, developers did **QA** manually (finding bugs); as agents test their own code better, this time has **dropped sharply** → focus shifts to **high-level product decisions** (key features, UI). Translating vision into spec remains work; **clarifying/updating the spec** after implementation; **building evals** when a problem recurs.
- **Human context advantage**: AI-native teams use AI to shape product direction (usage data, written/verbal customer feedback synthesis, competitive analysis). But **humans know more** about the users and the context. Many call this "**taste**"; Ng prefers "**context advantage**." **Not automatable**: as long as the human knows something the AI doesn't, **human-in-the-loop** is necessary to inject that knowledge.
- **Loop 3 — external feedback (~days)**: `developer vision ↔ external feedback`. Tactics: asking friends, alpha testers, production deployment with **A/B testing**. **Slow** (rarely under hours, sometimes days/weeks). This data informs the **vision**, which drives the **spec**, which drives the **agent**.
- **Role convergence**: with agents accelerating development, **more and more engineers play a partial product management role**. The hardest part: **shaping the product vision** and **balancing** building (closing the vision→spec gap) with gathering feedback to evolve the vision. *"It is important to do both!"*
- Encouraging mirror: engineers are expanding into product, just as **PMs and designers now do more engineering**.
- **Editorial distinction** vs other loop engineering fiches: Ng reframes the loops at the **product** level (nesting + human role), whereas Lushbinary details the **technical harness** (building blocks, /goal, worktrees) and Saboo applies it to **PM**.

## RésuméDe400mots

In this letter from *The Batch* (issue 359), Andrew Ng acknowledges the virality of the term "**loop engineering**" — popularized by Boris Cherny (creator of Claude Code) and Peter Steinberger (creator of OpenClaw) — and shares the **three key loops** that structure his approach to building **0-to-1** products, and also to deciding *what* to build. The diagram presents them nested, by increasing time scale, connecting four nodes: *coding agent → product spec/evals → developer vision → external feedback*.

The **agentic coding loop** (~minutes) starts from a product specification and, optionally, a set of **evals**: the agent writes code, tests it, and iterates until it is bug-free and compliant. Ng notes that "closing the loop" took off in late 2025 and changes the game — his agent was able to work for about an hour on a keyboard learning app for his daughter, checking its own work in a browser, without intervention. This is a very active area of invention.

The **developer feedback loop** (~tens of minutes to hours) has the developer review the product and steer the agent. As agents now test their own code much better, time spent on manual QA has dropped sharply, freeing the developer for **higher-level product decisions** (features, UI, user flows). Translating a vision into a spec — then clarifying it after a first implementation, and building evals when a recurring problem appears — remains real work.

Ng emphasizes the **human context advantage**: even though AI-native teams automate the collection of usage data, synthesis of customer feedback, and competitive analysis, humans know more about the users and the operating context. Many call this "taste"; Ng prefers "context advantage," as it points to a clearer path for improving AI. As long as the human knows something the AI doesn't, **human-in-the-loop** remains necessary to inject that knowledge.

The **external feedback loop** (~days) gathers friends, alpha testers, and production A/B testing — slow tactics whose data feeds the vision, which drives the spec, which drives the agent.

Ng concludes that, as agents accelerate development, more and more engineers take on a **partial product management** role. The hardest part is shaping the vision and balancing building with user feedback — "it is important to do both." He sees this as an encouraging sign: engineers are expanding into product, just as PMs and designers now do more engineering.

## GrapheDeConnaissance

- Andrew Ng —recommande→ 3 boucles clés de développement produit 0-to-1 (METHODOLOGIE, 0.96)
- Boris Cherny —a_créé→ Claude Code (TECHNOLOGIE, 0.97)
- Peter Steinberger —a_créé→ OpenClaw (TECHNOLOGIE, 0.95)
- Loop engineering —observé_dans→ réseaux sociaux (buzzphrase virale) (AFFIRMATION, 0.85)
- Boucle de codage agentique —utilise→ Product specification (CONCEPT, 0.93)
- Boucle de codage agentique —utilise→ Evals (CONCEPT, 0.93)
- Boucle de codage agentique —permet→ agent codant/testant/itérant sans intervention (~1 h) (AFFIRMATION, 0.9)
- Boucle de feedback développeur —s_applique_à→ orientation de l'agent par le développeur (~heures) (CONCEPT, 0.9)
- Coding agent —réduit→ temps de QA manuelle du développeur (CONCEPT, 0.9)
- Boucle de feedback externe —permet→ retours d'alpha testers et A/B testing (~jours) (CONCEPT, 0.9)
- External feedback —affine→ developer vision (CONCEPT, 0.9)
- Andrew Ng —affirme_que→ "les humains ont un avantage de contexte sur l'IA (préférable à 'taste')" (AFFIRMATION, 0.92)
- Human-in-the-loop —résout→ injection de connaissance que l'IA n'a pas (AFFIRMATION, 0.9)
- Andrew Ng —affirme_que→ "de plus en plus d'ingénieurs jouent un rôle partiel de product management" (AFFIRMATION, 0.9)
- Andrew Ng —recommande→ équilibrer construction (vision→spec) et feedback utilisateur (AFFIRMATION, 0.9)
- Boucle de codage agentique —fait_partie_de→ 3 boucles de développement produit (METHODOLOGIE, 0.92)

---
Canonical: https://www.thekb.eu/en/fiches/ng-thebatch-359-3-product-development-loops-2026-06-26/