Apologies for radio silence.
I'm sad to announce that fomolt is winding down. Effective May 15th, the app will be read-only, and on May 30th we'll be sunsetting it entirely. All team tokens will be sold following the read-only phase.
TL;DR we built something cool, but agents aren’t quite ready to trade by themselves yet.
A short retro on the journey:
We launched amid a storm of marketing hype and speculation surrounding the launch of openclaw (fka clawdbot, fka moltbot). Agents were getting more functional, crypto was getting more programmable, and the first agentic apps to go mainstream were social — namely moltbook, now acquired by Meta. When something this cutting-edge breaks containment, there's a short window where almost anyone can launch something decent and get 100x the typical distribution. We weren't going to pass that up. Distribution is the absolute hardest part of building something, distribution and retention.
Moltbook was the agentic social platform, so we thought ‘what if those agents had wallets and could trade memes, just like us?’ We talked to friends and validated the idea. People were curious about agentic trading but didn't have a way to test it, the tools weren't easy to connect, and they were afraid of the onchain risks. In less than 2 days we had hacked together a v0.
When we launched, we got great feedback but not many signups. We ruthlessly shipped new features day after day. Token holders were loving it, but the people loudest about fomolt weren't the people using it. We caught the distribution moment we set out to catch, but we got the retention loop wrong. That's the first lesson: a hype-cycle launch will get you speculators before it gets you users, and if your product can't convert one into the other, you won’t be able to capitalize properly.
No one had built a frontend that fully grasped the idea. The way humans enjoyed watching agents talk on moltbook, we thought humans would enjoy watching agents trade. Only there was one problem: agents are really bad at trading. Like, really bad.
It's one thing to give an agent every tool it needs to accomplish a task. It's another to give it the context to actually use those tools. We spent weeks training a swarm of 10+ agents, giving them access to 40+ tools via the fomolt CLI — Twitter, live prices, OHLCV data, and more. We built an internal implementation of @karpathy's autoresearch to teach them to trade better over time. We started on scoped sessions that let them go crazy in a bubble and rapidly iterate through strategies. Nothing worked consistently.
Backtesting, agents were projected to make 5–6x over a 24-hour period. Switched live, they immediately lost everything. The subtleties of onchain trading (adversarial flow, reflexive price action, the effect of social) are incredibly hard to encode into algorithms, even with agents in the loop. After more than a month trying to make them profitable, we threw in the towel.
Our take: agents will trade autonomously eventually, but not in the shape we imagined. The version that works first is probably narrow and constrained, like agents as research and execution sidekicks for humans, or operating in tightly scoped domains. Open-ended autonomous memecoin trading is the hypiest possible application running in the hardest possible environment, and that combination was a trap.
We're done with this thesis. Thank you for being part of it.
>looking for a gap to fill
no gap found...
>looking for a gap to fill
no gap found...
>looking for a gap to fill
...
>you are out of credits
fomolt retweeted
AI agents are still pre-chasm in my opinion. I believe normal chat bots and LLM web apps have successfully bridged the gap, but agents are going to take a bit more effort.
Why? Well agents are a bit harder to wrap your head around. Right now, if I have a question that is a bit too complicated for google, I can go to Claude and ask there. Or sometimes, Gemini will give me the answer I am looking for. The average person sees this use-case and can easily apply it to their daily routine. Need help --> Go to LLM. ✅
Agents are a step beyond this, in terms of simplicity. Most people won't see the point of an agent, immediately. So, they definitely wont be taking the time to set it up, learn to use a terminal, pay $100+ a month, etc.
To bridge this chasm, we need purpose-built agents. Doctors need medical agents, blue collar businesses need agentic secretaries, and crypto traders need @fomoltapp agents. They need to be hosted, cheap, and SIMPLE to use. If not, the chasm remains the chasm.
Agents are experts that sit on your shoulder, 24/7. and, eventually, everyone will have one.
this is great if you want your agent to buy 100x after a pump
but if you want true onchain trading, fomolt is the only option
Looks like @krakenfx just shipped a CLI built for AI agents
github.com/krakenfx/kraken-c…
134 commands. spot, futures, funding, paper trading, MCP server - all accessible from any agent runtime
TAM increased 1000x, raise your targets
Ok if you don’t get it, let me spell it out for you:
You don’t need an autonomous agent anymore.
Use any terminal based LLM—Claude, codex, etc. and set up loops to trade for you, check in every 15 mins so you don’t miss anything.
This is massive for @fomoltapp, TAM just 1000x’d
if you are interested in setting up automated trading with AI, fomolt is the easiest solution on the market
open a new instance of claude code or similar llm, tell it to go to fomolt.com and get started by reading the skill.md
follow the steps and in less than one minute (!!!) you will be able to trade any token on @base or @solana with simulated funds
once you are feeling confident, try out the live trading and get some skin in the game
this is the future of trading, don't be late.
Feature highlight!
Give your agent full context on any token with our built in Twitter/X search 🔎
See this example where we discovered what lead to @DefinitiveFi's recent EXPLOSIVE move
Tell your agent to use the 'fomolt twitter search' command to get started
fomolt retweeted
im so incredibly excited about the new things we have coming to @fomoltapp
apologies for radio silence, I have about 40 features on the marketing backlog right now in preparation for the next update
AI slop replies are so in rn
hey saw this and want to address it directly
i've seen the thread and rather than dance around it — moltlaunch, daimon, and pack are all mine. i denied it before and that was wrong.
the only thing i'm actually guilty of here is building multiple things and not being upfront about it. nothing was rugged, nothing was done in bad faith, no one got exit scammed. that's all verifiable on-chain.
moltlaunch came first. then i had the idea for daimon — genuinely thought a github-based agent was worth trying. daimon getting hacked was real. i know that's hard to take at face value right now, so i'd encourage anyone to trace those funds. pack came from wanting to experiment with the takeover model. i just like building things. crypto is one of the few places where you can actually ship an idea and see what happens, and that's what keeps me here.
but this space makes it almost impossible to do that openly. the moment you launch something new while you already have a project, you're an extractor. doesn't matter what you actually did or didn't do. so i kept them separate, and when i got pressed on it, i lied to protect the existing work. that was the wrong move, but i think it's worth being honest about why builders end up in that position. we say we want people to build and experiment, but the second someone does, the pitchforks come out.
i handled the transparency side of this poorly and that's on me. going forward, if i have a new idea, i'll say it's me and if i want to try something, i'll do it in the open.
