Exploring the echoes reverberating through time left by the technology of yesterday as we embrace the technology of tomorrow.
The workflow-orchestration tools of the early 2010s tried to solve a problem the LLM agents of the mid-2020s are now taking another swing at. The problem is the same. The substrate is different. Worth being honest about which parts of the previous attempt to bring forward and which to leave.
WWDC 2024 promised the most ambitious consumer AI rollout of the year. Most of what shipped is the easy part. The interesting part, the part that would have actually changed how iPhones feel, is still pending.
The headlines say GPUs are unobtainable. The headlines are about the hyperscalers. For everyone else, the picture in early 2025 is more interesting and more usable than it gets credit for.
MCP is the standard the integration layer of the AI stack converged on faster than anyone expected. Worth knowing what it actually is before you wire one in.
A framework I sketched a couple of years back: capture the priors that shape how someone reasons, ship the result as a portable artifact. Worth revisiting now that the substrate has landed.
OpenAI's o1 and o3-mini are the most public proof yet that scaling test-time compute beats scaling parameters at a category of problems. Worth being clear about which category.
Thirty days of letting an agent actually drive, write the code, run the tests, fix the failures, and then noticing the patterns in what changed. Not the productivity story. The discipline story.
A coding agent shipped by the model maker, built around the terminal rather than the editor. Worth working through what's actually different about that shape, and what it implies for the IDE category.
Every SaaS vendor shipped AI features last year. Most of them are billed as included. They are not free, they are baked into the seat price, and the seat price is moving.
Three frontier models on the table at the end of February, three different bets about what "frontier" means. The interesting comparison isn't who wins, it's who fits which workload.
Reasoning tokens have to live somewhere. The somewhere is your context window. Worth working through what the budget actually looks like once you stop pretending output is the only thing that costs.
Anthropic shipped a model that knows when to think. The interesting question isn't the benchmark numbers, it's what happens when reasoning becomes a default behavior instead of a separate product tier.
