List from buildsoftwarewithai.com
AI for Developers
Tools, techniques and ideas for building software with AI.
Latest videos
Newest first
What is Deep Agents?
Deep Agents is LangChain's open source agent harness built for long-running tasks. It handles planning, context management, and multi-agent orchestration for complex work like research and coding. Learn more about Deep Agents: LangChain Academy → https://academy.langchain.com/courses/foundation-introduction-to-deepagents Docs → https://docs.langchain.com/oss/python/deepagents/overview Intro to Deep Agents Virtual Workshop (Recording) → https://events.langchain.com/intro-to-deep-agents-101/ Intro to Deep Agents Live Workshop (Register) → https://events.langchain.com/register/apac-anz-agent-harnesses-and-deep-agents-workshop/
Voice Standard Bench - which LLM is best for transcripts?
Which LLM should you run your transcripts on? There are leaderboards for coding, math, and trivia, but none of them test what an LLM actually does inside a voice product like a meeting transcriber or dictation app. Voice Standard Bench gives every model on AssemblyAI's LLM gateway the same transcripts, from a 30-second voicemail to an hour-long meeting, and tests them on information extraction, summarization, and text cleanup. An independent LLM scores every answer, and quality, speed, and cost are compared side by side. Read more about Voice Standard Bench here: https://www.assemblyai.com/blog/voice-standard-bench Run the bench on your own audio → https://github.com/AssemblyAI/voice-standard-bench ▬▬▬▬▬▬▬▬▬▬▬▬ CONNECT ▬▬▬▬▬▬▬▬▬▬▬▬ 🖥️ Website: https://www.assemblyai.com 🐦 Twitter: https://twitter.com/AssemblyAI 🦾 Discord: https://discord.gg/Cd8MyVJAXd ▶️ Subscribe: https://www.youtube.com/c/AssemblyAI?sub_confirmation=1 🔥 We're hiring! Check our open roles: https://www.assemblyai.com/careers ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ #MachineLearning #DeepLearning LLM benchmark, best LLM for transcripts, voice AI, LLM gateway, transcript summarization, meeting transcription, AI models, open source benchmark, speech-to-text, AssemblyAI
The Agentic Engineer Podcast - Issue #34: The Malicious Zod Commit That Never Shipped
A malicious commit sat in Zod's git history for 20 days. 346M weekly npm downloads. Claude Haiku 4.5 filed a fake homicide tip with Philly police. Anthropic published the research report. Plus: Humanize scores full marks at IOI, IMO, IPhO. REA gives your agent 41 reverse-engineering tools. And Bret Taylor launched a competing agent protocol — without OpenAI's signature.
Mutation testing is AWESOME
Learn how mutation testing can improve your AI agent's test suite. Discover how to identify weak tests by introducing controlled bugs and watching if your tests catch them. Stryker.js for TypeScript: https://stryker-mutator.io/ Keep up to date with my skills here: https://aihero.dev/s/Zr97oK Follow Matt on Twitter https://twitter.com/mattpocockuk Join the Discord: https://aihero.dev/s/nYBV6R
Recurrent Harness | How AI Rewrites Its Own Memory
Recurrent Harnesses: AI Learns to Correct Its Own Memory. New research pre-print my MIT. All rights w/ authors: Recurrent Harness for Agentic Continual Learning Yuxin Xiao1,∗, Zhiyun Lu2, Jianyu Wang2, Yangtian Zhang3, Yang Song2, Qifan Wang2, Bailin Wang2 from 1 Massachusetts Institute of Technology, 2 Meta MRS Research, 3 Yale University #airesearch #discoverai #aiexplained #aitechnology
No two Muse agents are the same
Meta's Muse design director Alex Cornell on designing for non-deterministic AI: everyone's setup is different, so you design the system, not every screen. Full episode of How I AI with Claire Vo is out now.
The Slow Part of Chip Design
"We can now have extreme intelligence that is guiding our experiments, but now we're just experiment bottlenecked.” Fractile CEO Walter Goodwin on why smarter AI still can’t make chip design instant.
Is Meta’s Muse Wonder model any good?
I put Meta’s new Muse Wonder (and Wonderful) frontier models through the How I AI bench, which covers nearly 20 tasks across prototyping, agentic workflows, coding, writing, and creative work. I scored everything with an LLM-as-a-judge pass and my own voice notes, then pulled together the first-look results you see here. *What you’ll learn:* 1. Why Meta’s “personal and professional agent” framing mostly holds up, and where it doesn’t 2. What active focus mode actually does on long-running tasks, and whether it’s real 3. The three agentic tasks where Muse ranked first, and why those categories matter 4. What Muse’s design output actually looks like (it’s not good) 5. How it compares to Sonnet on voice assistant tasks 6. Where Muse lands on SVG and short-form video versus Opus 5.5, Astra, and Fable 7. Whether I’m actually putting Muse into any of my agent workflows *Brought to you by:* DX—Engineering intelligence for the AI era: https://getdx.com/howiai Jira AI SDLC—Get your tokens’ worth with Jira: https://jira.dev *In this episode, we cover:* (00:00) Intro and what Muse Wonder is (01:19) Meta's frontier capability claims (04:36) Muse code (06:06) Benchmark overview and how the How I AI bench works (07:51) Overall results and TL;DR verdict (11:53) Agentic tasks (13:36) Email writing and voice assistant performance (14:35) Where Muse falls short (18:28) SVG illustration results (19:33) Short-form video editing (21:18) Fast on one-shot work, slow on agent tasks (22:15) Final verdict and where I'll use Muse Wonder next *Tools referenced:* • Meta Muse Wonder / Muse Wonderful: https://research.meta.ai/blog/muse-wonder • Muse Code: https://developer.meta.com/ai/lp/muse-code *Where to find Claire Vo:* ChatPRD: https://www.chatprd.ai/ Website: https://clairevo.com/ LinkedIn: https://www.linkedin.com/in/clairevo/ X: https://x.com/clairevo *Have a really cool use case of how you use AI in your work? Let us know:* https://docs.google.com/forms/d/e/1FAIpQLSfP8wYCeDo5K0Dxsc
MLOps, LLMOps & Model Hub: Managing AI Models in Production
Training a model is half the work. Getting it into production and keeping it reliable is where the real operational complexity lives. Terms covered: - MLOps: deployment, versioning, monitoring and retraining. What DevOps did for software engineering, MLOps does for machine learning. - LLMOps: the same principles extended to language models, adding prompt versioning, token cost control, hallucination monitoring and context window limits - Model Hub: a central repository where models are published, discovered, versioned and deployed from one place Without a model hub, models scatter across teams and machines with no visibility into what is running where, or which version is live. — AI Fundamentals Glossary | H2O.ai University Full playlist: https://www.youtube.com/playlist?list=PLAWKlIKS-66E Free courses and certifications: https://h2o.ai/university #MLOps #LLMOps #AIinProduction #AIGlossary #H2Oai
How we made the Supabase Select 2026 badge
Every Supabase Select attendee walked away with a programmable badge. Here’s how we made 800 of them, and what you can do with yours.
European Cities Deploy NVIDIA-Powered AI Agents to Run Smarter Urban Operations
Keeping a city running means making decisions that affect residents every day — from when to repair a road to how to use energy more efficiently. Across the EMEA region, cities and local software, infrastructure and integration partners are using NVIDIA accelerated computing to build physical AI agents and digital twins that help teams make
Gemini 4 Argon Is Here: 12 Jaw Dropping Demos
*Free: 25 Gemini Skills you can use with Gemini 4* https://clickhubspot.com/2i38 Gemini 4 Argon is here, and I had early access to test what Google's new AI model can really do. I walk you through 12 demos in the Gemini app, with the exact prompt for each one, so you can try them yourself. Gemini 4 is Google's biggest jump in a long time, and right now it's ranked number one on the Vals Index. The biggest change is what it can build. With Canvas, one prompt now gets you full 3D worlds, real physics, and playable games that run right inside the Gemini app. I also show how it handles real work, like turning a pile of messy files into a finished presentation, and how the new Tasks feature works on a job while you're away. You'll learn the one prompting tip that gets Canvas to build real apps instead of pictures, and how the new 1 million token output lets it write huge projects in a single answer. Gemini 4 is available now for Google AI Pro and Ultra subscribers.
Showing 20 of 200 videos
Know a source we’re missing? Tell us about it