TIR 2025-06-28
by Gordon Weakliem
- Agentic Coding Recommendations -
- It’s this Blog’s 23rd Birthday - I have not been keeping up with my RSS at all. I’ve dedicated some time today, and Simon is a major part of my unreads. It’s amazing how he is so consistent, but he’s had 23 years to practice. In my origin story I mentioned that I actually started a weblog a couple of months before Simon, the difference is, I wasn’t consistent, so it’s still hard for me to get writing into a routine, and I don’t have what Don Box used to call the “blogging virtuous circle” of reading -> writing -> feedback. There’s a saying in Judo, “a black belt is a white belt who didn’t quit”. So Simon gets a whole section today.
- Quoting Kent Beck - “having a vision, being able to set milestones towards that vision, keeping track of a design to maintain or control the levels of complexity as you go forward. Those are hugely leveraged skills now compared to knowing where to put the ampersands and the stars and the brackets in Rust.” I think Kent absolutely nailed it.
- Quoting Andrew Ng - this linked into a gold mine of great stuff, from a summary of Mary Meeker’s latest Trends Report to a link to a course on Orchestrating Workflows for Gen AI. Seriously though, DeepLearning, where is the RSS feed?
- A bunch on sandboxing: browser sandboxing in Claude Apps, sandboxes in Cloudflare and Vercel, Gemini Canvas. In the last year, it’s gone from “where’s Google?” to Gemini quietly becoming a major contender.
- Project Vend: Can Claude Run a Small Shop? - I want a tungsten cube vending machine! Seriously, it’s a really interesting experiment. I do disagree with the assessment “Anthropic employees are not entirely typical customers” - I love that they tried to hack the machine. I’m positive that any real-world attempt to do something like this would result in asking the machine to stock marijuana products or alcohol within 24 hours. It’s also interesting that Claude both slept on profitable opportunities and offered products at a loss. Yet, “Although this might seem counterintuitive based on the bottom-line results, we think this experiment suggests that AI middle-managers are plausibly on the horizon”. This is actually one of the more fascinating things I’ve read in a while.
tags: [tir]
navigation