Today we're launching Manus Browser Operator. Any browser can now become an AI browser. One extension. No download. No new setup. Your browser already works. Your logins. Your sessions. Your habits. Now with the full power of Manus.

Lately i've been trying different LLMs for their OCR capabilities for handwritten tests. Gemini 3 Pro is definitely the best LLM for OCR right now. Previous best was Gemini 2.5 Pro. However, there are still many cases like this one which is not easily interpretable by an LLM.

Antigravity is a really cool name. Love it.

Ok here's the new flow: UI → Gemini 3.0 pro Code → Claude Sonnet 4.5 Plan → GPT 5.1 Just try this; remove all others.

Gemini 3 Pro-High is really good in Antigravity. That is until you hit the quota which is very fast.

I've been trying to find a way to describe the vibe difference between gpt-5.1-codex and Gemini 3 Pro. I think these code snippets highlight it pretty well

Google is so bad at being customer obsessed. It’s technology first, and when the hard part is done, move on to the next thing. Examples: - It’s impossible to upgrade from the Gemini app. - Google Workspace accounts can’t use YouTube Pro Family. - Minting an API key is the plot

LLMs make it easier to build custom software, but that alone does not justify replacing systems that already work. The real issue is whether the time and money spent will create a lasting competitive advantage. Rebuilding standard systems like CRM or HR tools rarely clears that

Deep down we all know Google trained its image generation Al using Google Photos but we just can't prove it.

As an AI expert who totally doesn't think LLMs are the path to AGI, but uses LLMs very avidly to get things done, here is my take on the utility of top LLMs currently: 1) If you don't know what you're doing in a certain domain, you can get interesting and valuable results, but

Claude: "I've defined phase one of your project to be completed during weeks 1-4 and updated the spec." Me: "Implement phase one now." *5 minutes pass* Claude: "Phase 1 is complete."

I get the hype around Claude Opus 4.5 but has anyone tried working in a team with some people using AI to generate code while others don't? It's such a mess.

Good lord, Claude Code now asks you questions before implementing.

I never quite understood why highly intelligent people say that SaaS is dead. As long as software exists, SaaS will exist.

Is this an Ad for Codex?

Next time your boss asks for a meeting on a weekend, show them this. If the European Union can wait out the weekend when a possible world war is going on, your boss most definitely can.

Genius

Codex is surprisingly good at code reviews

your company’s ci/cd pipeline.

This is the most important post of 2026 so far!! I don’t want to see anymore AI slop.