TOP LAUNCHES

GPT just got a speed upgrade

• GPT 5.3 Codex Spark is a smaller, ultra fast version of GPT 5.3 Codex built for real time coding inside the Codex app, CLI, and VS Code. It runs on Cerebras hardware with more than 1000 tokens per second, has a 128k context window, and is tuned for quick, targeted edits rather than long agent runs. It is rolling out as a research preview to ChatGPT Pro users with separate rate limits while OpenAI scales up capacity.

• Agent Bar is a native macOS menu bar app for running Claude Code without living in a terminal. Pick a project folder, spin up a session in one click, talk to it with your voice, and watch tool calls stream in as it works. You can approve or auto-approve actions, and keep an eye on token and dollar spend from the same little panel.

• SearchSeal tracks what the big AI models are saying about your brand so you are not guessing where you show up in their answers. It checks multiple AI platforms, watches how often you get recommended, and lets you monitor a bunch of brands in one place, which is especially useful if you are an agency juggling clients.

• Lindy Assistant connects to your email, calendar and work apps, then quietly takes over the boring parts. It filters and sorts messages, drafts replies that sound like you, lines up meetings and nudges you on follow ups so fewer things slip through the cracks.

• Happycapy runs Claude Code in a sandboxed workspace inside your browser and on your phone. You open a tab, get a visual desktop for agents, and let them handle coding, design, docs, and small workflows without setting up servers or local hardware. It is built by the original Trickle team and tries to make agent stuff usable for people who do not want to touch infra at all.

WHAT’S HOT

OpenClaw just picked a side

OpenAI has hired Peter Steinberger, the engineer behind OpenClaw, the scrappy agent framework that’s been quietly powering a lot of the “my bot actually does stuff” experiments. OpenClaw itself is not getting pulled into a corporate vault. It sticks around as an open source project in a new foundation run by Dave Morin, while Peter moves inside the OpenAI machine.

The spicy bit: Anthropic previously came down on OpenClaw over agent IP. Now the guy who built it is at OpenAI and the project has its own foundation behind it. That looks a lot less like a hobby project and a lot more like a growing ecosystem that one lab just lost the chance to be close to.

FROM THE FORUMS

Studying when AI does the homework

This thread starts from a simple worry: if AI can already explain anything, write your essays, debug your code, and even help solo founders ship products, what does studying actually mean in a few years?

The post points at CS enrollment dropping while AI degrees and “AI literacy” requirements explode, and asks whether school becomes mostly about learning the basics and how to use the tools, or if we need a completely different idea of what real learning looks like in an AI world.