💥 Issue 96

Spec-Driven Development 101, Future of Programming Languages, Interview with Boris Cherny, Modular Monoliths, Vector Search

Mar 05, 2026

Happy Thursday 👋 and a warm welcome to Tech Talks Weekly #96!

This issue includes over 120 new conference talks published in the past week together with 21 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!

“Backlog.md: The simplest project management tool for the AI Era by Alex Gavrilescu” from Voxxed Days Ticino 2026
Conference ⸱ +2k views ⸱ Feb 27, 2026 ⸱ 00h 41m 47s
tldw: AI coding agents like Claude, Codex, or Cursor make it easy to generate a lot of code fast but they also make work feel random and hard to review. This talk argues the missing piece is a simple task structure: keep the backlog as a plain md file, break work into small specs, and run a loop where the agent implements one scoped task at a time. The result is more repeatable output and clearer code diffs.
“Modular Monoliths: The Architecture That Scales w Your Domain - Omphile Matheolane - DDD Europe 2025” from DDD Europe 2025
Conference ⸱ +400 views ⸱ Mar 02, 2026 ⸱ 00h 42m 33s
tldw: Microservices are often treated as the default path to scale, but this talk argues many teams would ship faster with a modular monolith first. The idea is a single deployable app that is still split into real domain modules with clear boundaries using DDD tools like bounded contexts, context maps, and domain events. You get fewer distributed system problems, simpler deployments, and code that matches the business, while keeping the option to break out services later only when the boundaries and scaling needs are proven.
“The Past, Present and Future of Programming Languages by Kevlin Henney” from Voxxed Days Ticino 2026
Conference ⸱ +2k views ⸱ Feb 27, 2026 ⸱ 00h 49m 56s
tldw: Programming languages change way slower than the rest of software, and a lot of what gets marketed as new is just old ideas finally landing in mainstream. It looks back at how language ideas cycle, why adoption lags for decades, and what might nudge the next shifts, including open source culture and LLM-assisted coding.
“Building Claude Code with Boris Cherny” from The Pragmatic Engineer
Podcast ⸱ Mar 04, 2026 ⸱ 01h 37m 59s
tldl: Claude Code started as a side project and turned into a core internal coding tool. The talk gets specific about how it pulls the right context from huge codebases, how permissions and sandboxing keep it from doing anything sketchy, and what day to day usage looks like when the model writes a lot of the code. The bigger point is that engineers are still needed but the job shifts toward scoping work, steering the system, reviewing changes, and influencing product and design.
“Get The Most Out of Virtual Threads in Java 25: A Deep Dive by Christian Wörz” from Voxxed Days Ticino 2026
Conference ⸱ +400 views ⸱ Feb 27, 2026 ⸱ 00h 43m 54s
tldw: Java 25 virtual threads are now good enough for real apps without the weird edge cases that showed up in earlier releases, like pinning. This talk builds a small web app from scratch using virtual threads, StructuredTaskScope, and ScopedValue, showing how to run request work in parallel, cancel cleanly, and handle errors as a group.
Found something useful? Hit the ❤️ or leave a comment. Thank you 🙏
“Unmasking social debt: Silent threat to your team’s success • Kim van Wilgen • Devoxx Poland 2024” from Devoxx Poland 2024
Conference ⸱ +100 views ⸱ Mar 02, 2026 ⸱ 00h 41m 39s
tldw: Teams track technical debt but often ignore social debt: the trust gaps, unspoken pressure, and broken communication that quietly slow everything down. This talk argues that social debt feeds technical debt by making it harder to discuss code quality, challenge decisions, and collaborate on fixes. It suggests watching for warning signs like avoidance, low psychological safety, and information hoarding, then paying it down with clearer expectations, regular checkins, and habits that make feedback and disagreement safe.
“Vector Search Made Simple: Getting Started with OpenSearch for AI Applications - Dotan Horovits” from NDC London 2026
Conference ⸱ +800 views ⸱ Feb 26, 2026 ⸱ 00h 55m 25s
tldw: OpenSearch started as a search engine but now works as a vector database, too. The talk gives a quick intro to lexical search vs vector similarity search, then shows how OpenSearch can do both in one place using embeddings and KNN. It goes through common uses like semantic search, image search, and recommendations, and presents OpenSearch as a backend for RAG-style applications.
“Threads vs Coroutines — Why C++ Has Two Concurrency Models - Conor Spilsbury - CppCon 2025” from CppCon 2025
Conference ⸱ +6k views ⸱ Mar 03, 2026 ⸱ 00h 51m 58s
tldw: C++ has both threads and coroutines because they solve different problems, not because one replaces the other. Threads are OS scheduled and run in parallel, but thread creation, context switches, locks, and cache effects add real cost and complexity. Coroutines are language level tasks that suspend and resume cooperatively, so you can run lots of concurrent work on a small number of threads, especially for I/O bound code, but you still need threads underneath for actual parallel CPU work. The talk builds a mental model of what each maps to in the OS and hardware, then applies it to common cases like high fan out I/O, pipelines and CPU heavy work to help you pick the right tool.