YB
Over 9.7k subscribers
I hope all 8821 of you are having a great week so far 🔥
If you're enjoying my writing, please share Terminally Onchain with your friends in Crypto!
Today's post is brought to you by Splits!
Modern banking on Ethereum
Very soon, all organizations will need to set up onchain entities. Onchain businesses are more efficient and quicker to spin up than LLCs. The Splits' Teams product is perfectly positioned to onboard the next evolution of businesses.
Learn more how to start your first Onchain SMB in under 5 mins.
This past week, I was all over the place. First off, I want to apologize for missing Tuesday's deep dive. It was the first time in 4 months that I skipped a post. The truth is, I tried my best to get the piece together, but life caught up to me and I wasn't able to draft an article that I felt okay sending out.
Working for yourself is a double edged sword. I love the independence and freedom I get. But at times, I am pretty hard on myself as there's no one else to keep me accountable. So missing Tuesday's post definitely killed the mood but I'm over it and realized the more important thing is get something out today!
Part of me feels a bit burnt out so I'm excited to take a break for the holidays. It's been non-stop this year and I think my brain needs a few weeks off. With that being said, here's the holiday schedule:
Last post of 2024: Tuesday, December 18th
First post of 2025: Tuesday, January 7th
Last thing in terms of housekeeping...I actually wanted to ask a favor from all of you!
My goal for Terminally Onchain is to hit 10k subscribers by the end of 2024. The community is at 8.8k subs right now so not too far. If you find my content helpful and interesting, it would mean the world to me if you could share Terminally Onchain with your friends, on social media, or any crypto groups you're part of.
I really think 10k subs is doable if some of you can even share with just one person :)
Feel free to send them the tweet below or the link directly: terminallyonchain.xyz
With that being said, let's get into today's post.
I take pride in being very transparent about how I'm building this brand, my content creation strategy, what's on my mind, etc. So today's post is less of a technical breakdown and more of a ~casual brainstorm~ based on how I view the crypto x AI meta progressing over the next year.
Let's dive in.
Okay, so the last 10 days, I've been deeeep in the trenches of AI content. Think "SF AI cabal" scaling discussions.
I've been going on long walks and listening to the Dwarkesh Podcast, No Priors, Unsupervised Learning, etc. I've been having discussions with Claude to better understand the last 15 years of AI in order to have a better grasp on key papers, researchers, etc. I've been watching technical videos to get up to speed on the math and models. And I've been exploring different tools that I feel terribly late to.
You're probably wondering, YB why are you pivoting to AI...it's literally a bull market? Well, I'm not pivoting!! There is no more pivoting from AI to crypto back to AI back to crypto. As I mentioned in the availability cascade of Onchain AI back in October:
But to me, it's past the hype. Please excuse my dramatic next sentence but I had to...
In the past month, it feels like the twins, Cryptography and AI, who were separated at birth, finally have found each other and hugged after all these years of being estranged.
Since the launch of Truth Terminal two months ago (wow time flies!), I've been writing a ton of content related to agents and everything happening on X and Farcaster. At this point, if you've been reading my content in Q4, there's a good chance you're tired of hearing the word agent 😂
But here's the thing, I strongly believe that we'll soon hit a local top on the agent meta specifically. Recently, as I've been going through most of these projects, a good amount of new agents are mainly copycats and producing slop content. It's clear that we'll see a consolidation period where a few winners from this first batch (i.e. aixbt, zerebro) will be heads down cranking away on their roadmaps. And in parallel, agent infra teams such as ai16z and Virtuals will be prioritizing what's next based on their learnings from the past few months. Then, eventually there will be a breakthrough on a different kind of agent that has a unique value prop and will spark phase 2.
We saw the same pattern during defi summer and during the '21 NFT craziness. Remember Loot and Nouns a few months after BAYC? This is not to say that we won't see new agents in the choppy periods. I'm sure there will be some cool projects that will do well, but the key takeaway is that eventually there will be a shift in the larger meta.
As I mentioned in my post Don't midcurve your FAFO Fund, learn from Cobie! One of the things he always mentions is to pay attention to the metagame. The only thing we know for sure in a bull market is that attention is fickle. Overnight, it can feel like the entire timeline is talking about something else.
With that being said, just because I think agents might enter a quieter period does not mean the entire Onchain AI vertical will be forgotten! If anything, the opposite. I believe that it's only a matter of time before all the new entrants to crypto x AI will start dispersing across the stack.
Why? Because it's the infra that will enable new capabilities, frameworks, and design space for these agents.
This realization hit me last Friday when I sent out my deep dive on Vana and DLP Tokens. Remember, "agents" as we are seeing them right now are just one bubble in that entire ecosystem map above. Sure, things like middleware and agent orchestration are definitely less sexy and fun, but they're parts of the Onchain AI stack that are just as important.
The truth is, with all the research I've done so far on agents, I didn't really need to have a real understanding of AI. Agents are the interface layer of crypto x AI. This means that they're at the top of the stack, easy to understand, and consumer facing. In my opinion, people (including me) dissecting down all the things going on with agents aren't really speaking from first principles...rather we're just terminally online onchain so it's more so reporting on the trends itself. To be clear, this is important and I've learned a lot in terms of how things are changing, but I would be lying if I said that by studying Zerebro I now understand how fine tuning works, what open source data retrieval looks like, etc.
In fact, after sending out the post on Vana, I did a rapid fire learning session on other decentralized AI projects such as Olas, Pond, Gensyn, etc. And it was crystal clear to me that there was no way I would be an expert on the middleware and compute layers without going deeper into the AI side. The whitepapers were tough to get through!
Onchain AI is not just about agents. It's bringing all parts of the stack to an open source and decentralized model.
Since $GOAT launched, I would argue that the "AI Jailbreakers" got up to speed with crypto pretty quickly. Not just locking up tokens and figuring out the vibes of niche crypto communities, but even realizing that decentralized social will play a big part in making agents work.
My guess is that pretty soon it'll be time for us folks in crypto to get up to speed with at least the fundamentals of AI infra and how it all works. Those who realize this earlier will have a headstart when it feels like there is a sudden interest in the lower layers of the stack.
Most of you are subscribed to Terminally Onchain for my onchain strategy deep dives. Don't worry! I'm not stopping those haha. But, what I want all of you to realize is that in 2025, it'll be increasingly clear that you need to have knowledge of both verticals - crypto & AI - to really keep up with what's happening on the edges of the internet.
And I intend to be your guide through all of this, we're going to learn together!
Look I'm not saying we need to be s-tier AI researchers. But what we do need to do is stay up to date with the latest developments in both the centralized AI and the decentralized AI worlds.
"Apps inspire infrastructure, then that infrastructure inspires new apps" - Grossman & Grant
So, if we want to understand how the internet of intelligence is progressing, we need to understand not only smart wallets, stablecoins, and decentralized social BUT ALSO fine-tuning, compute, and training data.
I'm not sure about this just yet, but I'm even thinking of hiring an AI tutor to get me up to speed quickly. What do you think, any recs?
Regardless, I have an idea of how I can hit the ground running when I'm back in January.
So what's the plan? Two parts.
Decentralized AI
I'll of course keep you all updated on everything related to agents. But past that, it's clear to me that the next focus will be on the middleware layer:
Onchain model creators - going deeper into Hermes III, Pond, and Dither
Agentic Networks & Platforms - think Olas, Morpheus, Questflow, etc.
Distributed Training & Fine Tuning: Gensyn, TogetherAI, Lumino
User owned Data: Did Vana last week. Will look into Ocean Protocol & synthetic data tools as well
Additionally, I want to keep all of us up to date with the developer tooling that's also quickly being iterated on for onchain agents. Specifically, I'm referring to getting into the weeds of the Eliza repo, playing with the Virtuals GAME Sandbox, and experimenting with the Coinbase Onchain agent kit.
If agents are going to be the primary users of blockspace, then I also want to focus on what the bigger picture here is. For the decentralized AI community...
What's our north star?
What does this space look like in a year?
Who exactly are our competitors?
What's missing right now when compared to centralized AI?
What are the bottlenecks?
What metrics should we be tracking?
What does the "end state" of the operating system look like?
AI Research / Centralized AI
This section is going back to the "SF AI Cabal" I mentioned above. Most of the cutting edge developments in AI are happening at the offices of like 5-10 companies: OpenAI, Anthropic, Google, Meta, Microsoft, etc. It'll be important to know what these folks are actively thinking about AND understanding what the implications are for DecentralizedAI.
My strategy here is going to be a barbell approach:
On one end, stay up to date with the latest news, podcasts, releases, articles, etc.
And on the other end, read the most valuable papers and blog posts from the last 20 years (basically understanding the history to have a clear idea of what will happen next...something we do here often at Terminally Onchain)
For example, this week I read through Gwern's post on agents which was written back in 2017! It was a prescient read on Tool AIs (chatbots) vs Agents. Gwern makes the argument that agents will be the the more realistic path to super intelligence because they're economically more efficient and can learn at much faster rates by taking action.
And just yesterday, I was watching this podcast episode with Bill Gurley, Brad Gerstner, and Satya Nadella. It was fascinating to get the inside scoop from Satya on the early days of the AI wars. Back when Musk and Altman were just getting OpenAI started and trying to find a cloud partner. And a blog post from Dario on scaling that made AI's true potential 'click' for Satya. And the moment he realized that they were way behind Google because of the DeepMind acquisition.
In terms of format for the newsletter in 2025, I'm not 100% sure on this yet but the idea that's resonating with me the most is:
Tuesday: deep dives on the decentralized AI, agents, stablecoins, relevant zk and wallet infra, and Onchain AI developer tooling
Friday: curated list of learnings & news on AI research. My guess is throughout the week, I'll be sharing my learnings on Twitter and "learning in public" and then curating the best of that content in the newsletter for all of you
To me, the goal is to continue being terminally onchain and try my best to be artificially unintelligent (ha ha ha okay I'll stop).
I'd love to get any feedback on what you think of this. At the end of the day, this newsletter serves as a means for me to learn and provide high quality breakdowns and value for all of you. Feel free to respond to this e-mail or DM me on X 🤝
I'll end with the point that I think it's absolutely imperative (yea I said imperative) that we track the intersection of these two verticals as much as we can. I simply can't rationalize anything else being more important than this development in the next year.
That's all for today's deep dive.
See you all next Tuesday for the final TOC post of 2024!
Please make sure to share terminallyonchain.xyz with anyone to help me hit the 10k subs goal :)
74
