I really like making things, and have come to the unfortunate conclusion that I can't make everything. These ideas I'd desperately like to see built and will likely use. If you are working on any one of these, I'd like to expend whatever time, money, or social capital that I can to make sure it happens.
A LLM forward version of OpenSecrets that consumes all Federal updates (speeches on the floor, proposed bills and their changes, motions, etc) and feeds updates.
- It would really help to have all the activity summarized and changes flagged so we don't all have to constantly monitor what's going on. Even let people set custom alerts.
- Monetize against serving the updates as an API, and maybe increase the number of custom fuzzy alerts one can set.
- You could extend this to local government and make a killing for many large cap industries (PE, Real Estate, Energy, etc).
A new breed of LoRa wireless router that works well with very little power and is easy to deploy in annoying places. Think farms, factories, and solar farms.
- Traditional routers were built for offices and homes, not the outdoors. But there's really cool applications for IoT and Robotics that are dependent on great access to data.
- I've seen people solve this with 5G (which makes the device prohibitively expensive in many applications) or spending too much time optimizing software to need very little bandwidth. Things like humanoids can't work like this.
- The proliferation of Starlink makes this incredibly useful today. You should be able to spin up a tower + hub for these routers practically anywhere.
Small, $1000 humanoid-lite robots that I can use to run little experiments or play with. I would likely want several to train specific policies and see if I could get some of the learning to translate for other tasks.
- It doesn't need great arms or hands. I'm not trying to build another LeRobot or great actuation system.
- Components should likely be 3D printable so I can replace them on my own if needed
- Actuation doesn't need to be great, probably need to invest in good structure to manage CoG to increase the range of potential tasks.
Front + CRM (Twenty) + and AI Notetaker that I can quickly get set up with. Managing customers should be a solved problem by now.
- Responding to customer requests, generating pipeline and working through pipeline, and recording customer engagements should be the same product. Stop making me stitch together unreliable integrations or copy and paste
- The shared email client should be pretty simple; Front tried to do too much inside of one interface.
- Automatically piping notes and steps into the CRM is probably the only way to make the CRM usable.
Email inbox and phone/fax service that's built for LLMs and not for humans. Most of the world is really great at using email and phones!
- Probably served as something better than an MCP, maybe it's the client altogether.
- Most of the world already knows how to work with email and phone. This is probably the fastest way for LLMs to get good and doing useful things for non-technical people.
- Let the models have reliable access to communication. This is probably most useful for other humans, but eventually the models talk to each other over these services.
- Email servers make it really easy for humans to jump in when needed, if at all.
Solar is getting really cheap. Water transmission should start to get cheaper. Use cheap solar power directly onsite to a new type of DC optimized desalination plant.
- Sure, this is mostly useful on the west coast, but there's so much of California, Nevada, Arizona, and Utah that would be well served.
- Pump enough water, and you can start to terraform desserts into livable and farmable ecologies.
- More food, more housing, and cheaper access to water. Solve some of our biggest problems!