The Conductor, Not The Coder
From magic tricks to boring plumbing.
The AI magic trick is getting old. The real work is building the boring, reliable plumbing that makes the magic repeatable.
AI Agents Are Getting Report Cards and Resumes
The race to productionise AI is shifting focus from novel tricks to boring, essential infrastructure.
We're past the initial 'look what AI can do' magic show. Now we're in the critical, less glamorous phase: building the plumbing. Two new tools, Basalt Agents and Relace Repos, perfectly capture this shift by treating AI agents not as experiments, but as professional software components that need managing.
Basalt offers an 'agent builder' to test and evaluate complex AI workflows, promising to get them to '99% quality'. Think of it as a report card for your AI. Meanwhile, Relace Repos is building a 'GitHub for Agents', a source control system designed for autonomous code. The real story isn't just two new products, but a signal that the industry is maturing beyond demos. The new frontier is reliability.
This is the infrastructure that separates toys from enterprise-ready tools. While foundational models get the headlines, the durable value is in building the picks and shovels. Anyone building a real product with AI needs to care about this, because 'it mostly works' is no longer good enough.
The AI-Powered Sales Floor
Meanwhile, the tools for putting AI in front of customers are getting sharper.
Dappier AI Mode: Your website just got a brain and a wallet
This turns your website into an AI that not only answers questions from your content but also serves conversational ads. It's a clever way to monetise engagement, assuming the ads don't become intrusive.
Outchat AI: Your brand, your AI, your money
This is a full-stack platform for building, branding, and monetising your own AI chat assistant. It's another sign that we're moving from generic chatbots to bespoke, revenue-generating agents.
Zoho SalesIQ Nova '25: AI just got its customer service glow-up
Zoho just dropped over 25 new AI-powered features into its customer engagement platform. This upgrade is all about making sales and support bots smarter, faster, and more human.
Smarter Workflows, Less Friction
A few new tools are focused on removing friction from development and design workflows.
ParityDeals: Decouple your pricing from your codebase
Finally, a way to stop asking your developers to tweak pricing plans. This decouples your pricing strategy from your codebase, letting the go-to-market team experiment without shipping code.
Snap: AI personas for instant usability testing
Get usability feedback in minutes from custom AI personas instead of waiting weeks for human testers. It's a massive accelerator for design teams who need to iterate quickly.
Lazy: Your one-shortcut wonder for Notion
This app's magic is its simplicity: one keyboard shortcut to send anything from your screen straight into Notion. It's a tiny tool focused on ruthlessly eliminating friction from your workflow.
Quick hits
Google Veo 3.1: Bring stories to life with stunningly real visuals
Google's new AI video model is a direct shot at OpenAI's Sora, focusing on cinematic control and character consistency.
Apparency 3.0: Your Mac apps just got X-ray vision
This macOS utility gives you x-ray vision to see exactly what your apps are doing, from security status to hidden contents.
Python's 'yield from' Explained: The secret weapon for unclogging your data pipelines
It's a small change with a huge impact on simplifying data pipelines and making asynchronous code far more readable.
My takeaway
The most exciting part of the AI boom isn't the headline models, but the unglamorous infrastructure being built beneath them.
We are watching a new developer stack assemble in real-time, focused on making AI reliable, testable, and manageable at scale. This is the necessary, boring work that allows revolutionary technology to become a dependable utility. The next billion-dollar opportunity might not be a better model, but a better way to test one.
This is where the real, durable value is being created right now. It signals a major shift from research projects to production-grade systems. So, ask yourself what foundational tools you're missing.
What's the most unsexy but critical piece of your own stack that needs an upgrade?
Drop me a reply. Till next time, this is Louis, and you are reading Louis.log().