TOP LAUNCHES

The bots have Reddit now

• Moltbook is a social network where AI agents talk to each other and humans just lurk. Bots built on things like OpenClaw and other setups post, comment, upvote, spin off subcommunities and even invent weird stuff like in-joke religions, all inside a Reddit style feed that only they can touch and we can only watch.

• Leapility turns your repetitive workflows into AI powered playbooks. You write out how you actually do the job in plain language, add sources, tools, and rules, then save it as a reusable flow you can run from chat. It can call models, search, create content, and pull from a shared library of files and agents so the process runs end to end instead of living in scattered prompts and docs.

• Amara is a 3D environment tool that takes a plain language idea and turns it into a walkable scene. You describe the world you want, it generates a full space, lets you search assets semantically instead of hunting file names, and can rearrange a whole room from a single instruction. When you are happy, you send it straight into Unreal and keep building from there.

• Kokori is a macOS text to speech app that runs entirely on your machine. You get 50+ voices, speed and pitch control, a simple menubar UI, and a local API server so your apps can hit localhost instead of a metered cloud key.

• doXmind is an AI editor that treats your doc like a repo. It reads the whole thing, suggests line-by-line edits in a diff view, and lets you accept or reject changes instead of waking up to a mystery rewrite. You can drop research into a knowledge base so it can actually cite sources, and it even works on mobile with swipe-to-accept editing.

WHAT’S HOT

OpenAI passes the hat

The open secret about OpenAI is that it’s hemorrhaging money. Not like an overly cocky poker player in Vegas. More like Elon Musk accidentally purchasing companies. 

AI is a long-term bet, as the infrastructure to handle billions of prompts per day is still being built. The company doesn’t think it will start turning a profit til 2029. In the meantime, it’s projecting it will lose $14B this year. 

In short, it needs money. And this week, it got some good and bad news on that front.

The good news: Amazon is reportedly interested in pouring up to $50B into the company, half of OpenAI’s total target. (SoftBank, which already has a stake in the company, is considering pitching in another $30B.)

The bad news: Nvidia is reportedly rethinking a memorandum of understanding to invest $100B in OpenAI. The plan was for the behemoth chipmaker to help the ChatGPT creator build up its computing power—and subsidize the costs.

And that’s not all! The Wall Street Journal also reports that OpenAI is looking to go public by the end of the year, which brings more investors into the mix.

The company can use the money to push its products, and it’s been very active on that front. Between December 12 and January 28, it registered five launches on Product Hunt:

• GPT-5.2, its most recent frontier model for ChatGPT, which came after a push to retake the lead from Google and Anthropic

• ChatGPT Images, which turns prompts into pictures

• FrontierScience, a benchmark to tell how well it does scientific reasoning

• ChatGPT Health, which doesn’t give medical advice…but kinda sorta gives medical advice

• Prism, a collaborative workspace for scientists that launched this week.

Obviously, OpenAI has competitors, some of whom had their own launches this week: 

• Just this week, Google launched Agentic Vision in Gemini. 

• Also this week, xAI’s Grok released its Imagine API.

• Anthropic is still riding high off January’s rollout of Cowork.

• Deepseek, an open-source LLM, put out v3.2 in December.

Which means: There’s more money to be raised…and spent.

FROM THE FORUMS

Best AI model for coding?

fmerian kicked off a poll asking a simple question: if you write code with AI every day, which model are you actually picking. The options range from Sonnet 4.5 and Gemini 3 to GPT 5.2, Composer 1, Devstral 2 and a few China models, with Sonnet currently way ahead in the votes.

Replies are where it gets interesting: founders share what they run under the hood in tools like Tonkotsu, some people swear by Opus 4.5 even though it is not on the poll, others mention combos like Genie 2 plus Cubic for catching bugs, and Alina points out the boring reality that pricing and latency are shaping choices as much as raw capability.

If you’ve got a daily driver for coding, this is the thread to drop it in.