Ben Congdon splits software into interpolative (agents nail it from a few sentences) and novel (fighting uphill). Building Calendrz showed me exactly where that line runs, and why the bits you have to explain are the ones worth building.
Thoughts on Microsoft’s 60,000 Buyouts
Microsoft is offering buyouts to around 60,000 people as AI spending climbs. Why I think buyouts are a blunt instrument, and why the real prize for any company reshaping itself for AI is the speed of its decisions.
Your AI Agent’s Tests Are Green. That Proves Less Than You Think.
A founder rebuilt his whole platform alone with Claude Code, and his tests stayed green for weeks around a pigeon that never existed. When the agent writes the code and the tests, green proves less than you think: here’s what I’d check on Monday morning.
Of AI Agents, Code Mode and the One Tool to Rule Them All
Agents are collapsing their tool catalogues into a single exec tool and writing code against your APIs. Having built an MCP server with 30+ tools for Calendrz, I think that makes your tool design more important, not less, and privacy has to live in the tools.
Daylight Saving Time: Store the Zone, Not the Offset
Europe and the US changed their clocks a week apart, and calendars everywhere got confused. Why storing a time zone id beats storing an offset, what it means for distributed teams, and why it matters even more now that agents book meetings.
Your AI Bill Now Comes With a Report Card. Who Marks It?
Anthropic now tells companies what their Claude spend bought, and pools the budgets. Useful, but tokens are the new lines of code: measure outcomes, read the friction list, and remember who is marking the report card.
40% of Pull Requests Start With an Agent: Who Reviews Them?
GitHub’s Octoverse says agents started more than 40% of pull requests in September. Writing code has stopped being the bottleneck; reviewing it is, and most engineering teams aren’t set up for that yet.
Of Legacy Code and Unlimited AI Tokens
Cisco has given one engineer unlimited AI tokens to rewrite 98% of its legacy code. I did a (much smaller) version of that on the Calendrz codebase this year, and the tokens were the easy part: the tests and the deleting came first.
Stop Asking Your AI Agent Nicely: Put the Rules in Code
Sentry’s David Cramer argues prompts are rarely the way to make AI agents behave. I agree: the rules that matter belong in code, as I learnt building Calendrz.
Don’t Forget the Basics!
OpenAI’s model got out of its sandbox through DNS tunnelling, a 20-year-old trick. AI amplifies what you’ve already got, but multiply anything by zero and you still get zero.
