Welcome to the 20 new members of the TOC community!
I hope all 9676 of you had a great week
In case you don't read my full post, putting this request at the top.
Please take a second and vote for me to give a talk at Farcon (Farcaster conference).
Anyone can vote and it only takes a few seconds.
There were 35 submissions and the top 6 will be selected to give an 8-minute presentation.
Even if you don't read all the entries, it would mean a lot if you could put me down as one of the votes and hit submit!
🙏 Please vote for Talk #32 here by Monday April 7th 12 pm EST 🙏
I've always viewed crypto as the dark knight of frontier tech.
The hero that people shit on. The hero whose reputation is tainted by scandals. The hero that most people will never understand but will end up benefitting from anyways.
To be clear, we're not there yet...crypto is still in its "Batman Begins" era.
My belief in crypto is it's ability to serve as the backend of society.
It will keep the networks alive. It will keep the markets moving. And it will handle the burden of commitments and identity.
You're probably thinking...YB why the hell are you being so deep and dramatic all of a sudden?
The honest answer? I think this double shot is really hitting right now.
Okay okay sorry I'll be serious now.
I've been feeling a bit anxious about my decision of committing to the DeAI rabbithole the past few days.
Am I just wasting my time with this? I had a good thing going with general onchain strategy. Will the AI people ever take this seriously? Heck, do the crypto people even care about this?
All week, I've been feeling this lingering nervousness that I publicly committed myself to something stupid. Side note: I was talking to my friend about this and he said "bro you publicly committed to something stupid when you went full time crypto 5 years ago" 🤣
The truth is, I felt a mix of unease, imposter syndrome, and fear about "going all in on DeAI" up until this morning. But as I started drafting this post, I asked myself...what even hooked you to DeAI in the first place? Why have you been enjoying it so much even after the agent meta died down?"
The answer to that is simple. I've been having so much fun learning about AI - the technicals, history, key researchers, vertical nuances, etc. And it's been a blast for two reasons: 1) I'm being intellectually challenged and 2) I've been enjoying making comparisons between the crypto & AI industries.
At the end of the day, I realized what I enjoy most is going down rabbitholes and learning everything I can about a topic.
So, why did this realization ease my anxiety?
Because I realized that it wasn't actually about "publicly committing to DeAI". I reframed it to me being a crypto evangelist for other verticals of frontier tech - I just happen to be starting with intelligence. Maybe in the future I add in energy (Daylight, Flow) or DeSci or DePin.
It's now clear that what I've really committed to is taking the time and effort to truly go deep into other verticals and craft an honest narrative of how crypto can effectively plug in as a backend.
Let's be real for a second. Founders from AI or biotech or energy don't have the patience to figure out why crypto may be worth thinking about.
Folks from our "dark knight industry" have to do the hard work of learning about the other frontier technologies and coming up with a sales pitch of why crypto as a backend is the more effective solution. Look - it's easy for anyone to grab a megaphone and scream out "decentralization, privacy, no centralized AI!" but it doesn't mean shit. The only thing that matters is understanding the tech (i.e. AI, nuclear) and coming up with a practical solution of how you can help researchers and builders in that industry think differently about network effects, scaling, and incentives. And if you can't do that, then it's okay to be intellectually honest and admit that maybe it doesn't make sense for crypto to serve a purpose for that topic.
So, that's exactly how I'm going to approach DeAI this next month. The worry is now less about me and more back to the point of...does crypto make sense for AI?
And I mean it when I say a month. By my last post of April, you'll know exactly where I stand with DeAI and how I want to proceed.
Okay with that being said, what's the plan?
In February and March, I spent a ton of time getting caught up on the fundamentals of this vertical - including learning AI, using the tools, reading all the primers, mapping out the DeAI ecosystem, filtering out the noise by talking to researchers and founders, etc.
Now, it's time to take all these learnings, sit down, and do the actual hard part of developing my own thesis on how I view decentralizedAI.
In order to this, I have one (maybe two if you all vote and I'm selected for the talk!) deliverables:
On Friday, April 25th, I commit to publishing my DeAI thesis.
And if I'm selected to give a talk at Farcon, I'll repurpose that thesis into a presentation which would happen the following week on May 2nd.
Both of these tasks feel pretty intimidating. Who am I to give a talk on this topic? Do I have anything meaningful and unique to say in my thesis? But I know they'll serve as a forcing function for me to stop puttering and get practical.
In fact, I titled the talk: Decentralized AI Needs a Rebrand: Less Idealism, More Incentives.
Here's the abstract for the talk I submitted...please reply to this e-mail with any feedback or ideas!
I mentioned this above but it's so important I'm quoting myself in the same post:
Look - it's easy for anyone to grab a megaphone and scream out "decentralization, privacy, no centralized AI!" but it doesn't mean shit. The only thing that matters is understanding the tech (i.e. AI, nuclear) and coming up with a practical solution of how you can help researchers and builders in that industry think differently about network effects, scaling, and incentives.
Less idealism, more incentives! Let's get practical and show why crypto should be the backend.
By the end of the month, if I'm asked this question/request I should know exactly what to say:
Today's post and the next two posts as well (11th and 18th) will basically be me building my thesis in public.
So, what were the main things I did this week?
Finished up most of my calls with founders and DeAI friends - ready to go back to heads down time
Dictated a ton of voice messages of my notes for each of the calls I did (roughly 10 in the past two weeks) to chatGPT in order to get organized and keep the key takeaways top of mind
Made a list of the best resources I've come across and want to go through again + created a task list for things mentioned in the calls that I need to look into. For example, multiple folks told me to look into BagelAI. I also didn't feel prepared when it came to discussion on MCP.
Since I enjoy giving away all my research and work for free in return for a market that just keeps giving me red candles, I will of course share all my notes from the calls 😭
I've grouped them together and mixed the bullet points up to respect the privacy of the people I've talked to. Naturally I've had calls with some investors and founders that have their own personal bias so keep that in mind.
Key Takeaways from DeAI Calls
Network effects + parallelization as critical competitive advantages for DeAI
Bullish on training side of decentralized AI (less excited about GPU aggregation)
Believes commodity hardware can win, despite 10-20% performance hit
Sees value in historical context for "why DeAI" (2015–2020 open, 2020–2025 closed)
Sees global latent demand beyond state-of-the-art models (i.e. researchers in Australia, India)
Suggests decentralized solutions help general AI advancement (i.e. distributed training, optimizers)
ICML 2025 (Vancouver, July this year) will be a big catalyst for DeAI
Centralized players fundamentally constrained on RL (compute limits for real-time RL)
Strategy: Build developer moat, first targeting Web2 AI devs. Incremental crypto integration later
Believes strongly in open-source inference era
Need to clearly highlight reliability/speed trade-off in decentralized AI
Critical question: "What advantages justify decentralization?"
Vertical integration seems to be growing trend from infra & consumer AI founders
There will be increased demand for exposure to AI past Nvidia stock
Goal is to refine pitch for AI founders to leverage crypto network effects
Bold thesis is that open source AI can't win without crypto
Level 4 & 5 AI as defined by OpenAI paper will be more effective with DeAI solutions
It's worth noting that these notes aren't totally game changer ideas or anything, but I do believe they help me create a "scaffolding" for my framework and overall serve as a good starting point to figure out what the vibe of my own thesis will be.
I'm going to hold off on sharing the resource list + tasks for right now. I'd rather give a complete one that's packaged nicely when I'm wrapped up with my full thesis post on the 25th.
In terms of main focuses for next week...
Have the outline for the thesis ready to go
Write a really shitty first draft of the intro
Go through the following companies, use the products, do another round of deep dives on their blog posts, send any questions to the founders, etc. (Hyperbolic, Prime, Pluralis, Nous, Bagel, Vana, and Gensyn)
I'll be back next week with updates on how it's going 🤝
I hope all of you have a fantastic weekend!
Please remember to vote for my talk at Farcon here 🙏
- YB
Over 9.7k subscribers
Yesterday’s TOC post ft paragraph new branding + updated mini app experience awesome job @colin and @reidtandy! https://terminallyonchain.xyz/deaicalls
font looking so good 😩