What we're writing about.
Notes from the studio — development, design, AI, industry shifts, career growth, and life around the team.
Browse the archive.
Fable 5 Is Back — and Sonnet 5 Is the One You'll Reach For
June 30 was a two-headline day for Anthropic: the US cleared Claude Fable 5 to return, and Sonnet 5 shipped the same afternoon. One is the frontier you keep in your back pocket; the other is the near-Opus workhorse you'll actually run all day. Here's how they stack up.
GPT-5.6 Lands in Preview: Sol, Terra, and Luna
OpenAI is previewing GPT-5.6 as a three-model family — Sol, the new flagship; Terra, a cheaper everyday workhorse; and Luna, a fast budget tier. Sol sets a record on Terminal-Bench 2.1 and edges out Claude Mythos 5. Here's what shipped and how to get it.
Opus 4.7 to 4.8: The Upgrade That Fixed the Upgrade
Claude Opus 4.7 benchmarked well and felt worse — refusals, early give-ups, and a quiet token-cost creep. Here's what actually went wrong, how Opus 4.8 walked it back, and why we defected to Codex for the whole 4.7 run.
Have a project in mind?
The same team behind these articles builds production platforms every day. Tell us what you're working on.
Let's connect → [email protected]