Launched this week
SureThing.io
Autonomous agent that communicates results like a human
1K followers
Autonomous agent that communicates results like a human
1K followers
Everyone's running AI agents. Seldom hitting their business goals. AI isn't the bottleneck anymore. Humans are. SureThing is a General AI Agency. Paste any GitHub skill — it becomes a team you can @ anytime. One persistent memory across your COO, CMO, and CTO — zero silos. Agents that report up like humans. So you can finally run it like a CEO, not a debugger. With SureThing, now hit your business goals at inference speed.
RiteKit Company Logo API
The gap between starred repos and actually running them is real. The execution friction you're solving for is exactly what keeps most teams from shipping with these tools. Curious how you're handling the business context piece — that's usually where things get messy, especially when agents need to make decisions that actually matter to the business.
@osakasaul memory mechanism is actually our core differentiator. Most agent memory today is note-style. Extract, save, recall. Useful, but you still start from zero every session with better source material.
Ours is closer to procedural memory. We save the state of the work, not just what happened.
When you come back, your COO/CMO/CTO don't reconstruct from notes, they resume. Half-drafted email, pending vendor input, whatever was live in working memory still is.
You don't recall Tuesday. The agents are Tuesday.
The "Karpathy research agent / Garry Tan gstack" framing nails the real friction - discoverability is fine, but most starred repos rot, and "right there" usually means broken installs. Curious about the QA layer. When you turn a GitHub skill into a teammate, what catches dead repos vs. the working ones - runtime tests, community signal, or trust the maintainer? And once a skill drifts (a dep breaks, an API moves), does the employee silently fail or know to escalate?
@abhisheksinha123 Good points. Step one is standing on the shoulders of giants — importing external skills from the open ecosystem so every agent starts with proven capabilities, not from scratch. But the harder, longer game is what really matters: getting the agent to self-reflect, catch its own failures, and actually learn from them over time. That's the piece we're committed to investing in for the long run.
Love the "CEO not debugger" line.
Real question though: when the COO, CMO, and CTO agents share one memory, who wins when there's a conflict between sales goals and product reality? 😅
BTW, many congrats on the launch team!
SureThing.io
@boyuan_deng1 Haha reality always wins 😂 . Thanks man! 🙌
@boyuan_deng1 what a great question! Probably I will go for sales goals haha. Customer's willingness to pay explains everything. But do open for discussions.
@boyuan_deng1 LOL, I think it's "you", the CEO, tell them what you exactly want at that moment.😄
KnowU
I have like 30 starred repos and maybe 2 running lol. Congrats on shipping this, curious how the memory layer works across the COO/CMO/CTO agents — is it shared context or do they each have separate threads that sync?
@carlvert wow Calvert, you immediately locate our key competitiveness - memory. Most agent memory today is note-style. Extract, save, recall. Useful, but you still start from zero every session with better source material.
Ours is closer to procedural memory. We save the state of the work, not just what happened.
When you come back, your COO/CMO/CTO don't reconstruct from notes, they resume. Half-drafted email, pending vendor input, whatever was live in working memory still is.
You don't recall Tuesday. The agents are Tuesday.
@carlvert Great question. Each agent has its own memory and context, but can pull shared context from other agents, and call each other to collaborate on execution. So less three chatbots in parallel, more an actual team.
@carlvert Key point! How do you think about this? 😄
@SureThing.io best agent I've ever used so far. I'm using it for my linkedin post ideation to posting. how about adding personal skills?
@arjun_pansheria Thanks Arjun! Just tell your agent whats wrong, your digital extension will reflect itself and self-improve day by day.
@celine_yu thanks!
@arjun_pansheria Personal skills are very much the direction. SureThing auto-learns and evolves them at runtime as you use it, based on its own success and fail, and also your feedback.
And you can also bring your own skills if there are on other platform, or paste any open-source skill straight into your team.
From this angle, it's basically a machine for learning and internalizing skills.
@arjun_pansheria also, you can just share any best practices or skills with your agent, and let it learn them itself
@heyalicehan thanks you guys for update, I'm surely looking forward to it.
Love the chairman, not coder framing. How much business context can SureThing actually absorb before it starts making decisions?
@angelo_bram Honest answer: there's no fixed context bar. SureThing doesn't load your business upfront, it absorbs it through the work.
Mechanics: human-in-the-loop on every decision early. You approve, override, or redirect. Each becomes procedural state, not just notes. After enough cycles, the agent makes the call you would've made, and you stop reviewing the boring 80%.
@angelo_bram No limit. Try to connect apps which store or manage your business/data, SureThing can absorb connect contexts from them significantly!