Uploaded November 2026 | Updated November 2026, 4 days ago
Green tests donβt guarantee good AI-generated code π§ͺ
Matt Pocock explains how software fundamentals and three engineering guardrails can keep AI coding agents from filling your codebase with slop β‘
Whatβs inside:
πΈ How AI-generated tests can create false confidence β and how to make tests verify actual behavior instead of simply passing.
πΈ How automated checks, React architecture, and deeper modules help constrain AI-generated code and keep your codebase maintainable.
πΈ How to use a separate AI reviewer and more effective human code reviews to turn engineering standards and recurring feedback into a system that improves AI-generated code.
π€ Matt Pocock is an independent TypeScript educator and the creator of Total TypeScript, a comprehensive learning platform for developers. Today, he is building AI Hero, where he teaches developers how to apply real engineering principles to AI-assisted software development.
His work has already reached over 135,000 developers, with more than 13,000 trained through AI Hero cohorts. He also maintains an open-source collection of practical engineering skills for AI coding agents on GitHub, which has attracted over 323,000 stars.
π Explore his work: https://www.aihero.dev/
π Total TypeScript: totaltypescript.com
π GitHub: github.com/mattpocock
π Talk: Why Software Fundamentals Still Matter | Matt Pocock
This talk was recorded at React Advanced London 2026:
π reactadvanced.com/?utm_source=youtube&utm_medium=MattPocock
π Follow the link to watch the full version of all the conference talks, QnAβs with speakers and hands-on workshop recordings:
π gitnation.com/events/react-advanced-2026?utm_source=youtube&utm_medium=MattPocock
π Timestamps:
00:00 - Intro
02:04 - The AI Speed Trap: Why Your Software Factory Needs Brakes
03:17 - Three Guardrails That Keep AI Agents from Producing Slop
04:40 - Max Out Your Lint Rules: Give AI Agents Less Room to Make Bad Decisions
06:03 - Green Tests Can Lie: Three Ways AI-Generated Tests Create False Confidence
09:18 - Deep Modules, Better Architecture: Make Your Code Harder for AI to Test Badly
10:47 - React State Is Scattered Everywhere. AI Agents Struggle to Keep Up
12:19 - Turn Messy Frontend State into Reducers and State Machines
13:33 - Stop Trusting the Coding Agent with Your Standards: Use a Separate AI Reviewer
16:51 - Human Code Review Should Improve the System, Not Just Fix the Code
18:09 - Give AI Agents the Right Engineering Vocabulary to Make Better Decisions
19:42 - Turn Your GitHub PR Comments into Better Automated Checks and Reviews
20:00 - How React Knowledge Helps You Finally Get the Speed AI Promised
π The most honest conversations happen after the talk. Join the next conferences by GitNation to sync with the engineers actually shipping the future:
π« React Day Berlin - Dec 4 & 7, 2026 (Berlin & Online)
π https://reactday.berlin/?utm_source=youtube&utm_medium=MattPocock
π« AI Coding Summit Berlin - Dec 4 & 7, 2026 (Berlin & Online)
π https://reactday.berlin/?utm_source=youtube&utm_medium=MattPocock
π Check out all the upcoming events from GitNation:
π gitnation.com/events?utm_source=youtube&utm_medium=MattPocock
#GitNation #ReactAdvanced #AI #AIcoding #AICodingAgents #AIGeneratedCode #SoftwareEngineering #Programming #Coding #WebDevelopment #FrontendDevelopment #React #ReactJS #TypeScript #SoftwareArchitecture #CodeQuality #CodeReview #AutomatedTesting #SoftwareTesting #CleanCode #DeveloperProductivity #AIEngineering #AIForDevelopers #CodingAgents #Linting #StateMachines #React #ReactJS #ReactDevelopers #React19 #JavaScript #TypeScript #FrontendDevelopment #FrontendEngineering
Green tests donβt guarantee good AI-generated code π§ͺ
Matt Pocock explains how software fundamentals and three engineering guardrails can keep AI coding agents from filling your codebase with slop β‘
Whatβs inside:
πΈ How AI-generated tests can create false confidence β and how to make tests verify actual behavior instead of simply passing.
πΈ How automated checks, React architecture, and deeper modules help constrain AI-generated code and keep your codebase maintainable.
πΈ How to use a separate AI reviewer and more effective human code reviews to turn engineering standards and recurring feedback into a system that improves AI-generated code.
π€ Matt Pocock is an independent TypeScript educator and the creator of Total TypeScript, a comprehensive learning platform for developers. Today, he is building AI Hero, where he teaches developers how to apply real engineering principles to AI-assisted software development.
His work has already reached over 135,000 developers, with more than 13,000 trained through AI Hero cohorts. He also maintains an open-source collection of practical engineering skills for AI coding agents on GitHub, which has attracted over 323,000 stars.
π Explore his work: https://www.aihero.dev/
π Total TypeScript: totaltypescript.com
π GitHub: github.com/mattpocock
π Talk: Why Software Fundamentals Still Matter | Matt Pocock
This talk was recorded at React Advanced London 2026:
π reactadvanced.com/?utm_source=youtube&utm_medium=MattPocock
π Follow the link to watch the full version of all the conference talks, QnAβs with speakers and hands-on workshop recordings:
π gitnation.com/events/react-advanced-2026?utm_source=youtube&utm_medium=MattPocock
π Timestamps:
00:00 - Intro
02:04 - The AI Speed Trap: Why Your Software Factory Needs Brakes
03:17 - Three Guardrails That Keep AI Agents from Producing Slop
04:40 - Max Out Your Lint Rules: Give AI Agents Less Room to Make Bad Decisions
06:03 - Green Tests Can Lie: Three Ways AI-Generated Tests Create False Confidence
09:18 - Deep Modules, Better Architecture: Make Your Code Harder for AI to Test Badly
10:47 - React State Is Scattered Everywhere. AI Agents Struggle to Keep Up
12:19 - Turn Messy Frontend State into Reducers and State Machines
13:33 - Stop Trusting the Coding Agent with Your Standards: Use a Separate AI Reviewer
16:51 - Human Code Review Should Improve the System, Not Just Fix the Code
18:09 - Give AI Agents the Right Engineering Vocabulary to Make Better Decisions
19:42 - Turn Your GitHub PR Comments into Better Automated Checks and Reviews
20:00 - How React Knowledge Helps You Finally Get the Speed AI Promised
π The most honest conversations happen after the talk. Join the next conferences by GitNation to sync with the engineers actually shipping the future:
π« React Day Berlin - Dec 4 & 7, 2026 (Berlin & Online)
π https://reactday.berlin/?utm_source=youtube&utm_medium=MattPocock
π« AI Coding Summit Berlin - Dec 4 & 7, 2026 (Berlin & Online)
π https://reactday.berlin/?utm_source=youtube&utm_medium=MattPocock
π Check out all the upcoming events from GitNation:
π gitnation.com/events?utm_source=youtube&utm_medium=MattPocock
#GitNation #ReactAdvanced #AI #AIcoding #AICodingAgents #AIGeneratedCode #SoftwareEngineering #Programming #Coding #WebDevelopment #FrontendDevelopment #React #ReactJS #TypeScript #SoftwareArchitecture #CodeQuality #CodeReview #AutomatedTesting #SoftwareTesting #CleanCode #DeveloperProductivity #AIEngineering #AIForDevelopers #CodingAgents #Linting #StateMachines #React #ReactJS #ReactDevelopers #React19 #JavaScript #TypeScript #FrontendDevelopment #FrontendEngineering










