@boardyai building WithVoice, an AI fluency company helping professionals communicate clearly with AI, looking to meet enterprise buyers. would love to get Boardy Pro
Introducing Claude Opus 4.5: the best model in the world for coding, agents, and computer use.
Opus 4.5 is a step forward in what AI systems can do, and a preview of larger changes to how work gets done.
Bonus reading: Dario's challenge to the AI industry on the urgency of interp!
("Interpretability" being too much of a mouthful!)
https://t.co/REFmSZ2Gv7
Anthropic runs exit interviews for their AI models.
Just learned about this over coffee at their pop-up in Marylebone, early Sunday morning. Queued 20 minutes. 100% Worth it. 🧵
Thanks to @sammcallister and @whitneychn for the conversation and sharing more about Athropic's approachs.
Love how open Anthropic is about publishing their research and alignment processes.
https://t.co/B0hSv9MMzZ
We’re expanding Claude for Financial Services, with an Excel add-in, new connectors to real-time data and market analytics, and pre-built Agent Skills, including cash flow models and initiating coverage reports.
The counter dynamic to the AI model doing everything is that, at least in enterprise, bridging the AI models’ capabilities to the customer’s environment still requires a tremendous amount of long tail work.
The gap between an AI agent working for 90% or 95% of the solution and 100% is usually about 10X more work than most realize.
Getting access to the enterprise data, connecting to the enterprise workflows, delivering the change management that employees need to adopt the technology, handling the regulatory and compliance requirements of that industry, and so on all require some degree of highly dedicated focus in a domain.
There’s a strong analogy to vertical SaaS here actually. One would have thought that horizontal technologies could solve all problems in SaaS. But in fact there are endless very large companies that just hyper focus on a single domain, because that level of specialization is valued by the enterprise.
We will likely see the same play out with AI Agents in the enterprise as well. And in fact these domains will be far larger than traditional software categories because the TAM isn’t software, it’s work to be done.
Very fun debate, but I’m taking the other side.
New data insight: How does OpenAI allocate its compute?
OpenAI spent ~$7 billion on compute last year. Most of this went to R&D, meaning all research, experiments, and training.
Only a minority of this R&D compute went to the final training runs of released models.
I Am Voicepilled.
A major step forward in human–computer interaction won’t come from bigger models alone, but from how we talk to them, natively, with our voices.
More thoughts:
The https://t.co/uop4XiFxan microsite is live!
Think of it as a README for agents: a simple, open format for guiding coding agents.
It’s been awesome to work alongside peers from @AmpCode, @cursor_ai, @julesagent, @FactoryAI, and @roocode to bring this to life!