Also: human-in-the-loop theatre, MCP-on-Go pain, and Google Translate at 20.

Related to your interests

• Where AI Accountability Goes To Die When You Stick A ‘Human In The Loop’ And Call It ‘Governance’ – 🤖: “Stuart Lauchlan and HFS / Altimetrik’s Humans at the helm of AI study of 505 G2000 execs find enterprise AI is cost-driven, accountability-asymmetric (CIOs carry the blame, not the strategy), and the ‘human in the loop’ is mostly theatre – only 25% say humans would prevail in a disagreement, 18% have clear visibility into AI reasoning.”
• What Happened When We Treated AI Like an Engineering Teammate
• Everyone’s an Engineer Now – “Anthropic engineers are producing roughly 200% more code than they were a year ago, Cat noted. Today the main constraint is reviewing all that code to ensure it’s production-ready.”
• The Impact of AI-Generated Text on the Internet – “We find that by mid-2025, roughly 35% of newly published websites were classified as AI-generated or AI-assisted, up from zero before ChatGPT’s launch in late 2022.” // Also, chart on negative AI sentiment.
• The Slop Problem Isn’t What You Think – “A lot of AI writing is terrible. Soulless, repetitive, generic. But so is a lot of human writing. I’ve sat through vendor whitepapers with seven listed authors that would embarrass a school newspaper. We’re not actually angry about quality. We’re angry about dues. Did you suffer enough? Did you earn it?” // Good content is good content, and bad things are bad.
• Building a production MCP server in Go – ”The discovery endpoints above are straightforward to serve. The hard part is making the actual OAuth flow work end-to-end with your existing auth system.” // every time I try to make an MCP server, this is the problem.
• Cloudflare Announces Agent Memory, a Managed Persistent Memory Service for AI Agents – AI as a Service? Well, at least components of an AI stack as a service. Can you share it between multiple people? A shared context would be cool.
• Today I shipped 20 apps and a screensaver
• How To Blog – The Hemingway approach: “TLDR: Honesty is key. Write what you know and care about. What what you really think. Write original by being, doing, and thinking original.” // OP actually has much more.
• Failover Weekly: Ask the Hard Question – ”The best podcast interviews this week had one thing in common: the host didn’t let the guest off easy.”
• 20 fun facts to celebrate Google Translate turning 20 – Along with Maps, one of the best applications out there.
• Why are the Artemis II photos on Flickr?
• 🤖 Acetaminophen Quietly Beats Ibuprofen on Safety, and Drug Labels Won’t Tell You Why – Despite its narrow lethal-dose window, acetaminophen is generally safer than ibuprofen for most people because ibuprofen’s COX inhibition systemically damages stomach, heart, and kidneys. The FDA’s per-drug mandate and liability fears around comparative advice keep this fact off official labels.

Wastebook

• “[N]ever talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user’s query.” OpenAI’s Codex.
• “day-date” Seroter.
• “Matt Ray, but verify” SDT 570.
• “Here’s one we built earlier.” Blue Peter.

ICYMI

• 90 minutes of hell is other people
• The Enterprisification of Agents – Software Defined Talk #570 – “Google’s enterprise agent stack, what makes an agent an agent, and AI’s biggest critic. Plus, Coté wants better vision.”

Logoff

Nothing to report, really.

Want to subscribe to this newsletter and get it in your email? Do that here. You’ll just get this type of link and post round-up, not everything posted on the weblog.