How do we automate business analytics with Claude?
New blog post covering our best practices for skills, data foundations, and evaluations when building agents to perform data analysis:
https://t.co/mfEJMAQFBU
"Hulk Hogan is America" is one of the finest essays on the US I've come across. It just happens to take the form of an extended stand-up comedy clip by @JoshJohnson https://t.co/67kzvUHQyM
I don't have access to Mythos, but the cyber investigation capabilities of Codex+GPT-5.5 & Claude Code + Opus 4.7 are already very impressive when we used them as part of research into a new batch of suspicious websites this week. Can't wait to see what Mythos can do in this area
#NEWS 🚨: Artemis II crew experienced issues with Outlook this morning and had to ask ground crew for assistance
"We have two Microsoft Outlooks and neither one is working"
You can now enable Claude to use your computer to complete tasks.
It opens your apps, navigates your browser, fills in spreadsheets—anything you'd do sitting at your desk.
Research preview in Claude Cowork and Claude Code, macOS only.
Spoke about AI during #SGBudget2026. Many of us are familiar with its tools. But there's much more we can gain when we learn to harness them well. We'll make it easier to access AI learning pathways through SkillsFuture — so everyone can adapt and learn in a fast-changing world.
Happy New Year animated ASCII art shoot out between Opus 4.5, GPT 5.2 Thinking and 5.2 Pro.
Exact same prompt but v different results.
Which model did it better?
Here’s the effort by Opus 4.5
Continuing a silly tradition of the using the top AI models to generate animated ASCII art during the holiday season.
This time it took both @claudeai's Opus 4.5 and @OpenAI's GPT-5.2 Pro working together to get this design right. 🎄🎄🎄🎅🎅🎅
Merry X'mas folks!
Likewise, the Sora 2 failed again when I asked for a video showing the F1 Lego car transforming into the Lego X-wing version. No such sequence exists, and so it could not produce it.
Seems like a clear weakness in the Sora 2 model is its inability to "imagine" any physical sequences it has not seen before. In this case, I asked it to transform the Gundam robot into a plane. It failed, likely cos there aren't any sequences of this bot transforming into a plane
My personal benchmark for AGI: when a model can give me the instructions for rebuilding the Lego F1 car to the X-wing version. I’ll consider it AGI even if the model needs the original instruction manual for the F1 car
this is the conversation I have w OpenAI's reps in S'pore every time I meet them: How else are you upgrading Custom GPTs? Being able to pin a model to took way longer than expected. I don't want to have to wait as long for Custom GPTs to be able to use Deep Research or Agent mode
I'll note again that it seems nuts that, despite every AI lab launching a half-dozen new products, nobody is doing anything with GPTs, including OpenAI
When I talk to people at companies, this is still the way non-technical people share prompts on teams. No big change in 2 years