💥 Issue 93: Beyond the AI Hype, Gas Town with Steve Yegge, JDK24 Stream Gatherers, Advanced TypeScript

Run Any LLM on mobile, Supercharged testing w/ Playwright MCP

Feb 13, 2026

The most powerful Git client for Mac. (Sponsored)

Tower is the most powerful Git client for Mac and Windows. Over 100,000 developers are more productive with Tower.

Git Made Easy: Drag and Drop • Undo everything • A unique Conflict Wizard • File history
All of Git’s Power (And None of the Pain): Worktrees • Pull Requests • Single-line staging • Interactive Rebase • Submodules • Git LFS • Git-Flow • File History • Blame • Cherry-Pick
Software With Productivity in Mind: Quick actions • Single-click cloning • Automatic stashing & fetching • Quick open

Happy Friday 👋 and a warm welcome to Tech Talks Weekly #93!

Earlier this week, I published a compilation of the most-watched Java talks of 2025 which went viral. Make sure to check it out:

🏆 100 Most Watched Java Conference Talks Of 2025

Tech Talks Weekly

Feb 10

Read full story

This issue includes over 214 new conference talks published in the past week together with 86 new podcasts. That sounds like a lot, so make sure to check out both 🏆 Featured this week as well as 📈 Most-watched talks to find the ones you don’t want to miss.

Get ready and let’s jump right in!

🏆 Featured this week

Here are our top recommendations this week. Watch now or bookmark!

“Beyond the AI Hype: What’s Real, What’s Next - Richard Campbell - NDC London 2026” from NDC London 2026
Conference ⸱ +4k views ⸱ Feb 09, 2026 ⸱ 00h 51m 57s
tldw: AI is in a hype cycle, but some of the stuff is already paying off in real products. So what’s real, what’s exaggerated and what’s next?
“Gas Town, Beads, and the Rise of Agentic Development with Steve Yegge” from Software Engineering Daily
Podcast ⸱ Feb 12, 2026 ⸱ 01h 09m 52s
tldl: AI coding has moved past copilot-style autocomplete into tools that can refactor entire repositories, run multi step tasks and work across systems. Steve argues the hard part is no longer typing code but steering the work: keeping context straight, breaking goals into tasks, checking results and making sure humans and many agents stay aligned.
“Supercharged Testing: AI-Powered Workflows with Playwright + MCP - Debbie O’Brien” from NDC London 2026
Conference ⸱ +1k views ⸱ Feb 11, 2026 ⸱ 00h 48m 26s
tldw: This talk shows how to pair Playwright with the Playwright MCP server so an LLM can help write and maintain end to end tests. It runs through Playwright basics that matter in real suites like better locators, patterns that make tests less brittle, parallel runs and built in debugging. Then it demos AI assisted workflows like generating tests from a user flow, spotting missing coverage, repairing flaky failures and driving a live app with plain English commands.
“How I Tamed Claude - Emmz Rendle - NDC London 2026” from NDC London 2026
Conference ⸱ +6k views ⸱ Feb 09, 2026 ⸱ 01h 06m 11s
tldw: A CTO shares a workflow for using Claude Code as a coding agent without losing control of the codebase. The core idea is to treat the model like a junior developer: give it a real requirements doc, break work into small steps, review every change and keep a running project log so you and the agent can pick up later without relying on a huge context window. The talk also argues that junior devs still matter, but the job shifts toward clear specs, feedback loops and humans staying accountable for what is being shipped.
“GeeCON 2025: Marcin Chrost - Stream Gatherers - let’s get to know each other better!” from GeeCON 2025
Conference ⸱ <100 views ⸱ Feb 09, 2026 ⸱ 00h 51m 15s
tldw: JDK 24 adds Stream Gatherers as a permanent Java feature, giving the Stream API a long missing middle step between map and collect. The talk shows how built in gatherers can make stream code clearer for common tasks and then walks through writing custom gatherers.
“Indiana Jones and the Temple of Legacy Code - William Brander - NDC London 2026” from NDC London 2026
Conference ⸱ +1k views ⸱ Feb 11, 2026 ⸱ 00h 51m 15s
tldw: Legacy code is treated like an old temple: dangerous to poke at but packed with business rules you will not get back from a rewrite. The talk walks through ways to explore a mystery codebase safely, like using source control history to learn why things got weird, adding tests around the edges before changing internals and refactoring in small steps so you can roll back fast. It also calls out warning signs like cursed files with constant churn and hidden dependencies and suggests using them as maps for where to add coverage and boundaries first. The main point: your job is not to “clean it up” in one heroic pass, it is to learn how it works and modernize it without breaking your customers’ trust.
“Playing the long game - Sheena O’Connell - NDC London 2026” from NDC London 2026
Conference ⸱ +1k views ⸱ Feb 11, 2026 ⸱ 00h 46m 22s
tldw: LLM tools are everywhere now but most people either buy the marketing or get stuck arguing about hype. This talk maps out the common characters showing up in teams, from the “AI ate my job” dev to the “ban it all” gatekeeper and argues the useful path is somewhere in the middle. It says LLMs can help with fast iteration and prototypes but results get messy on big brownfield systems, so teams need the same old basics: tests, reviews, tight feedback loops, clear ownership and guardrails for where these tools are allowed to touch production code. The main concern is the long game: if juniors mostly prompt and paste, they do not build judgment. The talk advocates for keeping the real junior work around and building training paths now so they still have senior engineers a few years from now.
“The Fun Side of Advanced TypeScript: An Interactive Coding Session - Dante De Ruwe - NDC London 2026” from NDC London 2026
Conference ⸱ +900 views ⸱ Feb 11, 2026 ⸱ 00h 53m 47s
tldw: This one is a live demo of what TypeScript type system can do. It plays with infer, recursive and conditional types, type level math and even a full Wordle clone done entirely in types. The point is not that you should write production code like this, but that these tricks teach you what the compiler can do and give you a few new tools that can occasionally save the day.
“How to Run Any LLM On-Device With React Native by Szymon Rybczak | React Advanced London 2025” from React Advanced London 2025
Conference ⸱ +200 views ⸱ Feb 10, 2026 ⸱ 00h 26m 09s
tldw: Running LLMs on device changes a lot: lower latency, better privacy and offline support without round trips to a server. This talk shows how to do it in React Native using react-native-ai, a wrapper that hides the platform specific details behind a provider setup. It walks through hooking up engines like MLC LLM and Apple foundation models, then covers what recently got better: debugging tools, tool calling and agent style pipelines. The focus is on what works today and the tradeoffs you take on when the model lives inside the app.
“Why AI Projects Fail - Explore Top Reasons” from DevOps Podcasts
Podcast ⸱ Feb 12, 2026 ⸱ 00h 04m 18s
tldl: A lot of AI projects fail in one of three ways: they never ship, they ship then get abandoned, or they ship but nobody trusts or uses them. The talk argues the model is rarely the real problem. Most failures come from the surrounding system like messy data pipelines, unclear success metrics, weak monitoring and feedback loops, bad integration into existing workflows and missing ownership for upkeep after launch.