latest
I ran Claude Cowork against ChatGPT Work for a week — one clearly won
Someone's getting fired from my AI office.
This ChatGPT Codex feature is its best and worst at the same time
My unfinished projects finally have hope.
ChatGPT Study Mode is great, but these 5 prompts make it actually work as a tutor
ChatGPT's Study Mode is only as good as the prompts you give it.
6 file types you can upload to ChatGPT that are far more useful than PDFs
Uploading the original file is like giving ChatGPT the ingredients instead of a picture of dinner.
4 ChatGPT features that used to require a subscription — and are now free
Free ChatGPT can now handle jobs that once pushed you toward a paid plan. These four features could save you time and make it far more useful.
I taught ChatGPT, Claude, and Gemini the same lie — then tried to erase it
Gaslighting three chatbots for science
ChatGPT's Schedule feature does something most people don't know about — and it saves hours every week
I let ChatGPT babysit the pages I kept forgetting, and the results are mixed but promising.
I tested the top local LLMs against ChatGPT and Claude, and one absurd prompt revealed the gap
The local models weren't that far behind the frontier expensive cloud models.
ChatGPT just blocked its most controversial writing feature
There are already complaints that changes to the paid subscription are ruining writing for some users.
I stopped over-prompting ChatGPT after reading OpenAI's new guide, and my answers got better
"OpenAI's new prompting guide says shorter prompts reduce token usage while also improving answer quality.
Your Codex logs might have already killed your SSD without showing a single warning sign
AI wrote more logs than code.
3 things I do with ChatGPT Work that I couldn't with ordinary ChatGPT
I tested three ChatGPT Work tasks that regular ChatGPT left for me. Some worked better than expected, while others left me wondering why I bothered.
I built a website in ChatGPT in 11 seconds, and it wouldn't fake one feature
It writes convincing headlines, deploys without asking, and still won't fake a contact form.
I ran NVIDIA’s experimental speech model locally, and I was just not ready for how good it was
NVIDIA's PersonaPlex is fast, local, deeply impressive, and you can run it on just 8GB of VRAM.
When I gave Claude, ChatGPT, Gemini, and Gemma 4 an impossible ethical choice, their answers surprised me in different ways
AI models have developed a sense of morality — well, at least most of them have.
ChatGPT can now listen and talk at the same time, and that changes everything
ChatGPT can now listen and talk at the same time. Discover how GPT-Live-1 can make your chats feel more natural, useful, and easier to visit.
RAM prices are still crazy, so I had ChatGPT, Claude and Gemini build me a $1,500 PC — only one nailed it
RAM and SSD prices are up — so I let three AI models build the ultimate value gaming PC.
I taught ChatGPT and Gemini one thing Claude does by default, and the responses are better than ever
Two settings fields, a few minutes, and neither tool agrees with me on autopilot anymore.
I ran the same one-shot simulator prompt through Fable 5 and GPT-5.6 Sol, and the wrong model impressed me
I gave Claude Fable 5 and GPT-5.6 the same physics problem — the wrong model impressed me.
GPT-Live finally gives ChatGPT the one thing Gemini already had
ChatGPT's latest model should be a big step for its voice capabilities