Maybe you find this useful :)

A) Weird shit is happening, and its going to get weirder

• Steam, steel and infinite minds: Every era is shaped by its miracle material. Steel forged the Gilded Age. Semiconductors switched on the Digital Age. Now AI has arrived as infinite minds. Ivan Zhou (Notion CEO)

• Karpathy feeling behind, plus the ripple. The sands are shifting, Karpathy feels like he’ fallen behind (link)

• “AI will kill all the lawyers” (Spectator). A sobering read. “Last week we did an experiment, a kind of simulation. We took a real, recent and important case – a complex civil court appeal which I wrote, and it took me a day and a half. We redacted all identifying details, for anonymity and confidentiality, and we fed the same case to Grok Heavy AI. And then we asked it to do what I did. After some prompting, the end result was…’ He shakes his head. ‘Spectacular. Actually staggering. It did it in 30 seconds, and it was much better than mine. And remember, I am very good at this.”

B) Building with core AI

• LLMs under the hood: A BEAUTIFUL video showing how conceptually work. OMG it’s stunning. Welch Labs YouTube

  

• “You’re suddenly knee deep in chunking…” A pragmatic “ship something real fast” thread, aimed at avoiding infrastructure rabbit holes. X (formerly Twitter)

• State of agentic iOS engineering in 2026. A snapshot of how mobile dev workflows are mutating. And dev workflows in general. Dimillian

• Clémentine Fourrier: LLM Evaluation Guidebook v2. Good reminder that evals are the steering wheel, not an afterthought ngntipkolamrenang.twstalker.com

• Continual system prompt learning for code agents. People are starting to treat prompts like evolving infrastructure. X (formerly Twitter)

• tldraw: the fairies (agents) are amazing. Multi-agent coordination is a ‘fun’ puzzle to solve. X (formerly Twitter)

C) Greater productivity with AI

• How to prompt your Claude Code to keep it going from your phone (with VPN etc) granda.org+1

• React Grab CLI: Click in the UI, comment and automatically changes will be implemented agentically. Aiden Bai

• Prototyping isn’t thinking. debbiewidjaja.com

• Amp: thread maps. Clever feature to navigate your AI coding chats (in Amp). X (formerly Twitter)

<the end. enjoy your coffee, or matcha, or organic freeze thingy whatever>

p.s. Soz for being a couple weeks late. Holidays + sick children + sick wife + sick me.

Opinion

• Essay: The 2 problems of online consumption of AI generated content at scale: Signal erosion and verification difficulty. AND what you can do to counter them.

• Vertical AI is where it’s at: The next big wins are in narrow domains where data, workflows and evals are all specific and painful.

• Intro to LLMs: Karpathy’s talk still the best starting point. Here is a useful summary.

• “Thought does not require language. Language is an expression of thought. Intelligence requires thought more than language” - Yann Le Cunn

Productivity

• Compound engineering plug-in by Kieran Klaasen. With each feature you build, the plug-in: #Engineering

1. Documents patterns for the next feature

2. Creates reusable components that accelerate future work

3. Establishes conventions that reduce decision fatigue

4. Codifies knowledge that compounds across the team

• try out Claude Code’s frontend-design-skill to build UI without the purple gradients :D #FrontEnd

• n8n’s team built a Figma plugin that reads CSS tokens from Git and updates Figma variables in one click. No third party token tool, just their own automation #Design

• BugBot from Cursor is a low effort way to auto review PRs. Similar to Charlie and Codex review automations.

Novel use cases

• Clawdbot on X: Personal productivity bot by steipete. Powered by his duct-taped build of Warelay -- a CLI for sending/receiving WhatsApp via Twilio or WhatsApp Web.

• LLM audits as a service: Anthropic testing smart contract vulnerabilities - preliminary showing $4.6M in funds at risk with a PoC.

Evals

• Deep dive on LLM evals in The Pragmatic Engineer. Stop what you’re doing right now and read this!

  • p.s. I took the Maven course by Hamel and Shreya and cant recommend it highly enough!

Here’s the first weekly wrap of news. Happy bookmarking for later (that you won’t ever end up reading).

Opinion pieces

• Shipping is marketing, prioritising, building, testing and more. Practice shipping

• ProductTalk – “AI Changes Everything (And Nothing At All)”. The real AI skill for PMs is not prompting, it is resisting the urge to skip discovery just because shipping is cheap.

Novel designs

• Fast (sync) vs background (async) agents. Cursor’s Composer model is perfect for in the flow AI pair programming. GPT 5.1 Codex is perfect for delegated work

• Code heatmaps: Giving devs a bird’s eye map before they touch a file is quietly a 10x experience for onboarding and refactors.

• AI coming home to Sheets/Excel: Ramp Labs (yes the neobank) and Claude making moving into the tools finance and ops people already live in.

Deep dives

• Lessons from 6 months of using Claude Code to refactor 300k lines of code

• Vercel: “What we learned building agents”. The sweet spot of low-cognitive, high-repetition tasks.

• How Figma built human-centric AI evals for Figma Make, including design and functionality scoring.

“It’s easy to over-engineer your eval stack, your data set, some part of the quality loop — but it all depends on what users want to use your product for…. Building conviction on where you want to invest is the key part to get right.”

Learning

• LangGraphEssentials: Short course teaching LangGraph concepts like nodes, edges and state for building graph-style agents in Typescript.

• A quick start guide to MastraAI (LLM framework for Typescript). Get started in 5 minutes