Apple Sues OpenAI and Meta Ships Muse Spark 1.1
Apple says OpenAI built its hardware program on stolen trade secrets. OpenAI says it has no interest in Apple's secrets at all. That's the kind of week it was — and the kind of story Oscar and I actually enjoy making the call on, because the truth is somewhere in the gap between two press releases.
Here's how I read the week, and the two terms I think every decision-maker needed to have straight by Friday.
Signal or Noise
Apple sues OpenAI over alleged trade-secret theft
This is the one to watch, and not for the courtroom drama. It's a warning to every AI company hiring out of a competitor: talent doesn't show up empty-handed. If your onboarding doesn't deliberately separate what a new hire knows from what they were contractually not allowed to bring, your roadmap can quietly become someone else's evidence. Oscar made that point better than I did on the show. Signal.
Meta ships Muse Spark 1.1 and opens the Meta Model API
Meta opening its model API to developers matters more than the point-release. Signal, but a slow one — it only pays off when builders actually route real workloads through it.
GPT-5.6 across ChatGPT, Codex, and the API
Another frontier bump. I usually treat these as noise until they change what I can ship, and this one didn't move my stack. Noise for builders.
Grok 4.5 for coding and agentic work
Fine model, real progress. But "faster and cheaper and almost as good" is the story I care about, not "new leaderboard entry." Noise until the price does the talking.
Anthropic and OpenAI trade usage resets during the launch
Two labs resetting limits at each other during a launch week is theater. Noise.
No Jargon Required
Two terms an exec heard this week and probably couldn't define on the spot:
- Multi-agent orchestration — several AI workers coordinating on one job. It helps when the work genuinely splits into parallel pieces, and it multiplies failure when it doesn't. My honest advice: earn the second agent before you build five.
- Model routing — how a product quietly switches among fast, cheap, and capable models per request. This is where the real margin is hiding, and most teams paying premium tokens for every keystroke haven't turned it on yet.
That last point is the through-line for me: the frontier model is becoming the part you use least. Get orchestration and routing right and you don't need one genius doing everything — you need a smart conductor and a cheap, fast bench.
Watch or listen to the full episode below.
