HeyJaredTrilogy AI Center of Excellence
AI Center of Excellence turns advanced AI into results, guiding teams from concept to scale and sharing practical best practices every day.
- Indexed issues, last 90 days
- 15
- Latest publication
- Oct 1, 2026
- Audience
- Checking…
- Earliest in this view
- Jul 23, 2026
Latest issues
OpenAI dot: Why Would I Start a New Chat? (opens the original)
Read excerpt
OpenAI announced dot at DevDay on September 29, 2026, with access rolling out gradually. It’s an always-on AI agent in ChatGPT that can work across connected apps, keep working between conversations, and come back with results or questions. It has its own cloud computer, so some work can continue even when your laptop is off. That’s the pitch. I wanted to see what it actually felt like to use.I started using dot with a fairly obvious question: what is actually different here from regular ChatGPT
Jev's Open Rivals Test the Decision Model Idea (opens the original)
Read excerpt
On September 15, TypeSafe introduced Jev as a model that takes state and bounded questions and returns typed, probability-bearing decisions. Agent code can use those decisions to route work. A valid answer can still send a task down the wrong branch.Laya, Kev, AnyJev, and DiffusionGemma structured reads now return the same kind of typed, probability-bearing answer through different computational routes. Their value in an agent depends on how much work each route can accept at a measured error ra
A Charter Is Not a Control (opens the original)
Read excerpt
On June 12, 2026, a US export-control directive required Anthropic to cut off Claude Fable 5 and Mythos 5 for foreign nationals. Inside the US and outside it. Including its own employees.Real-time nationality verification is not a thing you stand up over a weekend, so Anthropic turned both models off for everybody, which wasn't really a choice, it was their only option.Thanks for reading Trilogy AI Center of Excellence! Subscribe for free to receive new posts and support my work.<for
Dario Amodei’s AI slowdown plan leaves a governance gap (opens the original)
Read excerpt
His plan specifies lab access and reporting rights. My earlier article also covered government threat claims and emergency restrictions.Dario Amodei commits Anthropic to continuing external inspection in We Must Pace the Frontier. His proposed reviewers would scrutinize lab safety while government threat claims and emergency restrictions remain outside their described mandate.Nearly three months earlier, in my
Astra reasoning effort: Does xhigh use fewer tokens than medium? (opens the original)
Read excerpt
Astra used more tokens as we increased reasoning effort across 24 runs on two engineering tasks. At xhigh, average token use reached 1.76 times medium for refactoring and 2.52 times for mobile redesign.The result runs against a claim circulating among Codex users: asking Astra to think harder could make it cheaper to use. The explanation is plausible. If more reasoning prevents a mistake or avoids another request, it can save the agent from processing a long conversation again. Enough avoided wo
Publishing over time
Last 90 days. Choose a month to open its work.
Recurring subjects
Named in the text we hold. One piece can cover several.
Audience
No verified audience measurement yet.
About this data
Counts cover the work we have indexed. Tone needs enough text and a confident classification. Excerpts and episode notes are not full articles or transcripts.
Identity or attribution wrong? Suggest a correction.