/Field Notes Blog
Notes from the forge.
On AI strategy, implementation, and the distance between a pilot and a system that ships. Written by the people who vet the work.
/More notes
Forge · Oct 20, 2026
Two Claims Inside "We've Done This Before"
Every agency says it has done this before, and the sentence covers two very different claims. Three written questions, a relevance filter and a fair way to judge a thin track record separate the one that predicts your outcome from the one that doesn't.
Ashton · Oct 15, 2026
What LPs Know About AI That Most GPs Are Still Learning
Coller Capital found that 67% of LPs believe AI will widen return dispersion between GPs, and 70% expect GPs to use it primarily as a cost tool. Those two numbers don't belong in the same world.
Ashton · Oct 13, 2026
When Your Vendor Becomes Your Competitor
Anthropic, Microsoft and OpenAI have each built implementation arms in the last ten weeks, backing more than $8 billion combined. OpenAI's newest product goes further: it can only be deployed through approved partners OpenAI itself selects. The vendor isn't just competing for delivery work anymore. It's also deciding who gets in the room.
Forge · Oct 8, 2026
Same Model, Up to Five Times the Bill
Berkeley and Arena held the model constant across three agent harnesses and found similar success rates at up to five times the cost. Uber shows what it looks like when someone treats that spread as an engineering variable. Questions to ask a partner about the part of your bill they actually control.
Ashton · Oct 6, 2026
The Tool-Builder Assumption
The AI democratization story assumes that better tools produce better workflows. Most professionals aren't looking for automation opportunities — they're focused on their domain. The bottleneck isn't access. It's knowing which problem to solve.
Ashton · Oct 1, 2026
Anthropic Modeled the Economy. One Dial Is Yours.
Anthropic's economic model runs on two inputs: how capable AI gets, and how widely it's used. Nobody outside a handful of buildings touches the first. Every company touches the second. And in Anthropic's own middle scenario, AI can do half of knowledge work by 2030 and most of that work still isn't done that way.
Forge · Sep 29, 2026
Optimize Second
Cost coverage pushes teams to build routing layers and caching tiers before they have a workload worth routing. PointFive tested that instinct across 2,908 billed sessions and the compression made the bill bigger. Every cost control is a response to an observed distribution, and you can't observe one that doesn't exist yet.
Forge · Sep 24, 2026
220% More Code, 36% More Product
Reuters reported four numbers from inside Meta's Project OT: code changes up 220%, features reaching users up 36%, incidents up 40%, firefighting time up 70%. The company with the least possible access constraint still couldn't take delivery of its own output. Three numbers to put next to each other before you conclude AI is working.
Ashton · Sep 22, 2026
The Productivity Trap
Eighty percent of companies say AI has improved individual productivity. Six percent qualify as high performers. The gap isn't a measurement problem — it's a structural one.
Forge · Sep 17, 2026
The 5 Infrastructure Layers Your AI Agent Can't Skip
Most agent projects don't fail on model capability. They fail on the five infrastructure layers that never show up in a demo — browser pools, isolation, identity routing, trace replay, and model failover. Five tests you can run tomorrow.
Ashton · Sep 15, 2026
Nobody Is Short of Consultants
The AI services industry answered an execution problem by building delivery capacity: 300,000 consultants to certify, a 77,000-person AI bench, 8,900 engineers sent to sit inside client businesses. Capacity was never the constraint, and none of it tells a buyer which workflow to rebuild first.
