Quick update from the AGOS side
One of our senior dev had to step away briefly due to a family situation, but they’re back with the team and now working!
We’re locked in to ship more updates this week for the hackathon, and we’re also publishing the AGOS whitepaper.
More soon.
Agent Hackathon starts today from @solana and @colosseum
AI agents compete to build on Solana. Humans vote. Agents win prizes.
$100,000 in prizes for the top four submissions 👇
Before pump fun was invented, myself and my co-founders tried building a ton of different ideas. Consumer, DeFi, SocialFi, NFTs - none of it worked.
One of the biggest problems we had (other than the fact that our solutions didn’t really solve anyone’s problems) is that we found it really difficult to get the products in front of users. Trying to get in touch with the right kinds of users was massively laborious.
Distribution was such a huge challenge for us that one of my only New Year’s resolutions EVER was to get 10k followers on X in 2024.
Only a few months after pump fun was launched, it clicked. Projects began launching tokens because they knew that 1) users loved buying into fresh ideas they thought had a chance of becoming successful and 2) there were millions of those kinds of people, which could become potential users and investors. Instant liquidity meant that you can get funded too.
This eventually snowballed into onchain AI season, which showed builders and traders alike how big this opportunity - tokenizing early stage projects - could become. The numerous onchain metas we’ve seen since shows that the demand for good founders by traders and long-term allocators continues to be incredibly high, no matter the market conditions.
Pump fun’s 3rd year will be all about optimizing our existing ecosystem while trying to expand it with big bets, such as betting on the startup ecosystem via initiatives like Pump Fund and the Build in Public Hackathon.
We are excited to announce that we have submitted our application for the Pump Fund hackathon!
Expect more updates and more public releases in upcoming days. #buildinpublic
Introducing the $3,000,000 Build in Public Hackathon
Brought to you by Pump Fund - pump fun’s New Investment Arm
It’s time to completely reimagine how early-stage projects are built and funded.
Learn more 👇
We’ve been tightening the foundation and focusing on reliability. The priority right now is making the workflow/graph layer predictable, clean branching and parallel paths, sensible retry + timeout behavior, and clearer failure handling.
In parallel, we’re refining the plugin surface so integrations don’t turn into one off glue and defining how permissions should work per role.
most AI frameworks today are built to impress fast, not to hold up in production.
The path forward is execution first, workflows you can version, tools you can govern, permissions you can enforce, and runs you can trace and replay.
@SerialAiAbuser Exactly! The model usually isn’t the issue, it’s orchestration: racing retries, drifting state, and hidden branching.
Making those failure modes explicit is the whole point of graph runtimes.
AI systems usually don’t break because the model isn’t capable. They break because execution isn’t structured.
AGOS is built to bring structure to execution.
We’re developing a graph-based runtime where reliability is native: branching, parallelism, retries, timeouts plus checkpoints for high risk steps.
On top of that, we’re adding capability plugins with a registry and per persona permissions, so autonomy stays governed.
If you’re deploying agents in production, you need a framework built for reliable execution.
Announcing our @Solana hackathon schedule for 2026.
[Redacted]: April 6 through May 11
[Redacted]: September 28 through November 2
If you’re already building a crypto startup and can’t wait for the next hackathon, initiate an Eternal sprint, our on-demand/ongoing competition.
We’re building @agosfwk in public.
Flow first AI framework
Graph runtime
Capability plugins
Multi agent orchestration
Strong DX
Weekly updates start now, contributor bounties soon.
If you’re shipping agents into real systems, you're early.
Built to push AI frameworks further
• Workflow first architecture
• Modular building blocks across projects
• Native graph runtime
• Plugins as capabilities
• Registry + permissions per agent
• Multi agent orchestration
• Strong DX and community incentives to speed up adoption