AI research is exploding.
If you want to know what's going on in AI, you need to be reading papers. NeurIPS (the top conference in AI) alone accepted 5,272 papers in 2025. No one is reading all of that.
Someone needs to build the discovery layer that scales with this growth.
We're building FiveW to do just that.
The Problem Compounds
It's not just volume. Three main problems:
- Volume: You can't read 500 papers a week in your subfield. Nobody can.
- Relevance: Of those 500, maybe 5 actually matter to your work. Which 5?
- Trust: You don't know if a paper is good until you read it—or see who else is reading it.
The current workarounds are broken:
- Twitter: The algorithm is unfocused. Papers get buried under memes and hot takes.
- Lab Slack: Siloed to your immediate team. Not searchable. Meant for communication, not discovery.
- Following specific people: Dependent on an algorithm you don't control. You don't know who you don't know.
The cost of missing an important paper? Your research becomes irrelevant. You waste compute on outdated methods. You get scooped.
Research is the New Literacy
Here's what's changed: anyone can do research now.
Access to compute? Cloud GPUs, Colab, Lambda. Access to pretrained models? Hugging Face, OpenAI API, open weights everywhere. Access to knowledge? Papers are free. YouTube explainers. Twitter threads. The barriers that kept research locked in academia are crumbling.
Look at the people shipping meaningful work without traditional credentials. Alec Radford never finished his PhD before writing the GPT paper. ditto for Chris Olah and undergrad before founding Anthropic. George Hotz built comma.ai by reading papers and implementing. The pattern is clear: agency can overcome a lack of credentials.
OpenAI and Anthropic are launching research residency and fellowship programs. They're explicitly saying no prior ML/AI experience required. Net-new researchers entering the field every cohort.
This is an inflection point. Research isn't just for PhDs anymore. It's a core skill for anyone building in AI.
The Research-Builder
The best companies in AI right now are research-to-product pipelines.
OpenAI began as a research lab and turned frontier research into a mass-market product in ChatGPT. Midjourney evolved from a research-driven Discord into the dominant image generation product by staying at the frontier. MSL invested tens of billions this fall to recruit top AI researchers and translate that work into deployed models. Their moats are engineering as well as research translation.
SOTA moves fast. If you're building in AI and not on top of the latest research, you're building on yesterday's foundations.
Our thesis: all builders will need to be research-literate to stay relevant. The line between researcher and engineer is blurring. The engineer who can read a paper on Monday and ship an implementation on Tuesday will win.
What We're Building
FiveW is the discovery layer for AI research.
A personalized feed tuned to your interests. Social signals showing what top researchers are reading. Collections to organize and share with your team. A simple, seamless way to find and digest the papers that matter to you.
We're starting as a layer on arXiv. But as research continues to explode—and it will—we're building the infrastructure to scale with it.
The bottleneck isn't creation. It's discovery. We're fixing that.