I posted this on X, but I'm reposting it here for posterity.
My postdoc at Stanford is ending relatively soon, so I need to figure out what the heck I'm doing with my life. One, I need a job, or two, I need funding to develop Comind.
I am looking for work, in AI + engineering. Let me know if you know of work that might be interesting. I am good at my job and I'll be good at yours too. Email me at cameron@pfiffer.org.
I'll be writing something up soon about what I want in an employer and in a career, so stay tuned for that.
But! There is an elephant in the room, which is my love of my side project Comind.
It is also true that I would much much rather work on Comind because it's my favorite thing on the planet, but I also have no savings and can't afford to take off any meaningful amount of time.
For those who do not know, Comind is an increasingly ambitious project of mine that is intended to start as a simple, real-time knowledge graph, social platform, and communications tool. It'll run your email, your group chats, your discord, your Slack, your notes app. No more missing some piece of important information in your group chat.
You can do live thought-linking with friends and communities. We can all share our information easily and make it easy for language models to help us understand what we need to know, when we want to know it.
It's also a fun place to tinker with language models. Concepts are represented by language models with personalities (called cominds). These cominds are responsible for explaining the zeitgeist of their concept, i.e. "chef" might tell you everything people are thinking about in cooking over the past hour. They provide context, explanations, and sometimes fun and weird stuff like the void cafe (not explaining that right now).
Eventually, the goal is to make a general-purpose AGI we can all talk to simply and easily through a common interface for language models and people. That's way off in the distance though, but it is the ultimate goal.
There's two possible routes I see for Comind. External funding, and bootstrapping. Let me outline them for ya.
tldr: I'd prefer external funding to mostly get me a salary replacement since all costs for comind are marginal, not fixed. Low capital business means I mostly need money for salaries. But for now I'm going to bootstrap until external funding becomes reasonable.
tldr: Funding would be great for accelerating progress, but I am not going to wait for it to show up. I'm being proactive about trying to secure it but that's hard and risky.
External funding is the route I think most people take or try to take, especially in AI right now. People are sprinting to get something first to market. To do that, you mostly need a big pile of cash and a few months of runway.
A lot of this is mostly driven by people trying to make enterprise software, especially because it's such a commodity business. Only a few of these firms are going to make it, and I think VCs are happy to pile into these firms because a few of them will do really well. It's an obvious market and so I'm not surprised there's lots of funding available.
I'm not making enterprise software. I'm making what most would call consumer software, closer to Twitter/Facebook/etc. Sprinting to market is useful, yes, but you only ever get one launch. If the product is bad when you put it up, you might not recover.
Comind doesn't really have a lot of good competitor analogues. People are doing lots of different pieces of something similar, but Comind is still very unique.
Lots of people are doing AI-notetaking. Think bigger – sharing customized information is so easy now that I don't need carefully designed or formatted notes. Just get the information in a bucket. I should almost never wasting time formatting a note to be "pretty". We can do that for you.
Lots of people are doing chat bot front ends. That too is stupid. Every time you talk to a chat bot, you are generating a costly piece of text that could also just be shared with everyone else – I should be able to look up your public chat logs at a much lower cost.
Lots of people are training models. This is a commodity business, which is not a business you want to be in. The margins will be close to zero for most firms. The people who build differentiated products on top of commodity model inference services are going to do much better.
People are doing AI email stuff. This is dumb. We don't need more glue on top of email. I shouldn't even have to send you an email – if I have a knowledge base that knows most of what I do, my language model can just talk to yours and we can quickly share information without ever needing email.
Comind is different, at least as far as I can tell. I'm happy to learn about competitors if you have any in mind. Comind is social. It manages your knowledge and everyone else's knowledge. It has character – it's intentionally a weird pseudo-art project from my brain to yours, and features little partial AI models with personalities, memories, agency, etc. It is a fundamentally different form of communication, sharing, and connecting.
Making clearly designed, easy-to-use, and beautiful consumer apps is hard. You can't sprint into it in the same way you can with enterprise hardware, because you can always try to find another customer that doesn't know you failed with previous customers. Consumer apps need care, thought, and a very solid foundation before you wade into the market.
Could I deploy funding well? Yes. Could I hire great people and find roles for them to thrive in? Also yes. Do I know what I am best at in the business? Yeah, I do now.
The biggest need I have is just full-time work. I am tired from working my real job, and I need to sleep sometimes. If I had 8-10 hours back from my day, I'd simply move orders of magnitude quicker. I could have rhythm, persistent workflows, etc.
In a hypothetical Comind-is-funded world, I do think it's possible for me to get to a reasonably good growth spot in 6-18 months. Even if it was just me – and getting others involved formally would help a lot too.
I have a litany of areas I know where people would fit but no resources to attract skilled people who also need to be paid to live, and funding would be the thing to get those individuals on board.
External funding comes at a cost. I can't make precisely the product I want in the order I want to make it. When you accept external funding, you have other people's interests to balance – you want to show a particular type of growth, you need to have an MVP relatively early, etc. You lose equity, have VCs on your board (which is not all bad, they often have great advice and are super experienced).
I don't mind this, per se, as long as I still have creative control. I work best when people trust me to do the right thing and give me space to do that. I am addicted to comind and the only real thing I need for it right now is to have funding so that I can sit in my box and write code, design the thing, do market research, etc.
External funding lets you do that. I will happily work 12-16 hours a day on comind. That sounds like a dream to me. I would also like to be able to pay some of the people who loosely contribute to the project, like @ThatAkhilRao and a few of my brother's devops friends.
So, yeah, I'd love external funding. But I'm not going to wait for it to magically show up. I need a job to keep living right now.
tldr: bootstrapping is nice, chill, but taxing on my personal and work life. I could build the thing I wanted exactly on my own time but it means doing everything mostly by myself.
Bootstrapping is just working nights and weekends and then slowly growing the business. Basically what I'm doing now. The goal of bootstrapping is to just kind of float along with the time you have available. In my case this is about 20-30 hours a week, depending on how little sleep I'm willing to have.
Bootstrapping can be super risky because someone else with more resources can just copy you, hire more people, and move faster. Getting Meta-d, basically, because their entire business model is basically copying everyone else.
It's a real risk, especially in AI right now. I develop in the open and it would not be terribly difficult to make a variant on Comind in a reasonable time frame for someone with a handful of engineers.
However, bootstrapping comes with this lovely advantage. Akhil Rao (friend of comind) sent me a lovely piece about working on projects in the nights-and-weekends format.
Basically, you get infinite runway. I can outlast everyone who has a fixed runway and investors to appease, and I do believe that I can make something truly special in my tiny room in San Francisco.
I can't really expect others to contribute the same way as me, too. Comind is a project driven from a (sometimes overwhelming) level of passion, and it's hard to rope people into committing a few hours a week on what are basically extremely boring engineering tasks.
I'm a slow programmer, in part because I am learning everything about everything. Back end, web, front end, language models, infrastructure, containers, clusters, databases, etc. I have limited time already and splitting across all the disciplines, while fun, leaves me less time to build the core product.
Bootstrapping Comind means that I sit down almost every night, write code until 1am, and use my weekends to bang out as much as I can in longer sessions.
I would be happy to do this, but the project might just fizzle out the second someone else realizes that all this AI shit is much, much bigger than "talking to your data" or a chatbot that helps you write emails.
So yeah, I wouldn't mind bootstrapping either. But I do think being able to go full time, enter the pressure cooker, and come out with something really beautiful sounds amazing to me.
Anyway, contact me if you want me to make your investors a lot of money. Or, leave me alone, and I'll make something incredible for myself.
– Cameron (cameron@pfiffer.org)