TOP LAUNCHES

Five AI tools you may have missed

• Hopper — Built by Hypercubic (YC F25), whose CEO left Apple ML after shipping to 200M+ devices to solve the most overlooked problem in software: the billions of lines of COBOL still running every major bank. An agentic dev environment for mainframe/z/OS that inspects data, edits code, writes JCL scripts, and diagnoses failures autonomously. Cursor for mainframes.

• MiniCPM-V 4.6 — From OpenBMB, the Tsinghua University NLP lab MIT Technology Review named one of the Chinese AI teams to watch after DeepSeek. A 1.3 billion parameter vision-language model that runs entirely on-device on mobile hardware. No cloud, no API call, no upload.

• Khaos Brain — A local memory system for AI agents that turns task experience, preferences, and workflow lessons into Git-versioned cards the agent retrieves before working and updates after. A Sleep/Dream/Architect maintenance loop keeps memory transparent and reviewable rather than a black box.

• display.dev — One CLI command gives your agent-generated HTML a permanent URL behind company SSO (Google, Microsoft, or OTP), with inline comments for iteration. For teams who want to share what their agents produced without emailing files around.

• Jotform Claude App — Jotform, one of the most widely used form builders in the world, shipped a Claude integration that lets you build, edit, and analyze forms directly inside Claude. Enterprise form infrastructure as a Claude skill.

SPONSORED BY

Are you really still typing?

Full disclosure: Wispr Flow is the AI dictation tool most of us at Product Hunt (use we still have a few holdout typers, what romantics). Hold a key, talk, and clean text drops straight into whatever app you're already in — Slack, email, Notion, your IDE, wherever your cursor lives. No switching windows. No copy-paste ritual. Just say the thing – yes, you can whisper it – and even your most run-on sentences will be turned into polished writing at 4x the speed of typing.

WHAT’S HOT

Yet another default

Another week, another default model for OpenAI. The company this week released a new model, GPT-5.5 Instant, and made it the default for ChatGPT. According to OpenAI, it’s “smarter, clearer, and more personalized” than its predecessor.

What’s that mean in practical terms? It hallucinates less, can check your math homework for you, and knows how to sort though past prompts and your files to get the answer you are looking for.

Some Product Hunt users are giving it a spin. Here are their early takes:

• Anusuya takes OpenAI at its word but is a bit surprised: “Making the smarter model the default is a bold move; most companies charge more for better.”

• Moh writes that the big upgrade is memory sources so users can see where context is coming from: “Every personalized AI system has the problem of ‘why did it say that’ — and the answer is usually buried in something the user can't inspect. Making personal context visible and editable is a harder UX problem than it looks.”

• Bogdan says that it feels more “capable” now, but he’s noticed something odd: “It seems like temporary chats now start mixing up with the main projects/chats and referencing them more. It's either just starting to do that, or it was way less obvious about it before.”