I created a totally free 'challenge' course on LinkedIn to encourage folks to try different general AI techniques. 10 challenges during weekdays, 10 mins each, help on Fridays. Although it's already started you can catch up. Get it while it's free:

Most people look at price per million tokens, but today, the real costs come during inference. the reasoning part of the tools. So Maybe we should pay more attention to The cost to run real benchmarks since it shows more realistic differences.

Join millions who have switched to Grok.

RT @JamesMontemagno: We launched a new podcast for @code, the VS Code Insiders Podcast keeping you up to date with the latest and greatest….

As a design/developer, one of the problems is that AI Agents are so decent at design that most developers are perfectly happy. But the designer in me sees those as misalignments to brands. We need to see AI that shows design options before coding things.

GitHub Copilot’s New AI Powers Will Change How You Code

Please call it the Emergency Medical Helper (until you get a hologram).

@sama Looks like it's a problem with having my VPN on, but most normies won't know to look at that. Still, it shouldn't have to rewrite the HTML and go straight to Canvas.

@sama The HTML is actually accurate, but for some reason preview is just showing a blank. It shouldn't need to retype it. Canvas is broken at this point, I've seen this several times.

@sama Meanwhile in canvas, after giving me decent HTML, this is what I see in preview.

@sama

This is a problem for ChatGPT @sama I give it a simple PDF extraction prompt and it does a great job of generating the HTML, but it should have pulled up Canvas to do this. When I request canvas mode, it regenerates the HTML it already did and then fails to show anything.

Also, you know what would really be great @OpenAI , letting us use the OSS models in ChatGPT with the tooling. Those are great models. Maybe it could be like ordering from the secret menu at a restaurant. a small setting to flick this on. Love the open models.

This is the best feature about ChatGPT-5 nobody's talking about, an emergent assistive attitude that suggests really smart improvements to your chats. More of that please @OpenAI

It'd be better if @OpenAI just gave us turbo mode, a no reasoning model without the router slowing things down. Or even a slider to control the thinking and reasoning. One model, controls for pros.

Alright, I took ChatGPT 5 through it's paces from a real practical view, there's a lot to like and some places where it misses the mark. It's clearly a superior model, but not in that way where when you've been eating dollar ramen all your life and someone takes you to a good

AI's Next Big Leap: Making AI Thinking Visible

Have you tried the styles popup in ChatGPT. Basically it adds predefined prompts to an image that style it a certain way. Would be more powerful if you could create these. Interesting to note the structure of these prompts. Here's the cyberpunk one. Create an image in a

GPT OSS is the most consequential model we've every had and it doesn't matter that it doesn't perform as well as other models. It's ridiculously cheap and extremely resourceful. It's easily going to dominate the Open Source AI market.

Why is it that nothing can beat Midjourney for creative images. The quality of the HD images are insane.