@levie Seeing this already with my Claude Code experiments in non-coding things (I call it "Claude Compadre" because it's fun) Example: implementing a system to help me process my dump of captured tasks, having specialized agents for breaking down tasks into visible actions, higher

@levie Yes lots of maturity needed before the new AI products are enterprise friendly.

@levie Exactly 👏 The real unlock isn’t just one powerful agent, but a swarm working in parallel. Startups that design AI-first workflows from day one will feel like they’re operating at 10x speed compared to legacy companies.

@levie Yes this is going to be huge in a few years and common place - won’t be able to imagine life before it.

@levie Exactly. Patterns will have to change, from workflows to auth! https://substack.com/home/post...

@levie What is your guidance with best practices for AI first workflows for new startups? Has anyone done a good job at taking a snapshot of current tools and methodologies and providing some recommendations?

@levie Interesting. Perhaps future AI agents could even manage other AI agents, creating a hierarchy of productivity.

@levie Curious what use cases beyond coding you think will unlock mainstream adoption first? - Sydney

@levie For the last century the big question was 'How can one person do more?' - better tools, more automation, because labour was expensive. Now it’s flipped - we add more agents with cost of intelligence coming down. What will then help scale is figuring out new ways for agents to

@levie Oh my... Social media is not yet ready for subagents or swarms and swarm theory :-O

@levie Most existing companies are mediocre at best. Which means you are correct: The very phrase "change management" embodies anathemas to creativity/long-term success. If one actually looks at what's required, many parallels to effective (yet rare) product engineering practices that

@levie Agreed, AI agents are adapting to human workflows, and humans are also adjusting so agents can perform better. I think we’ll find the sweet spot soon, though it’s more likely to come from startups or lean organizations.

@levie Could I get your thoughts on this? https://community.openai.com/t...

@levie Micro agents

@levie yes super exciting times ahead for both enterprise and startups. time to remake industry and economies.

@levie The leverage from swarms of AI agents is undeniable. The open question is how we calibrate the balance between machine autonomy and human authority as this paradigm scales across every area of knowledge work. Curious how you see governance fitting into this future of workflows.

@levie I think designing these workflows will be the critical part of AI adoption. With one AI/Agent/... the concept is pretty easy to grasp and integrate but once you make it a network the possibilities (and failure modes to be fair) multiply and organizations are not ready for it.

@levie Absolutely! We’ve been building Templafy since 2014, a document generation solution for enterprises. We didn’t realise that the orchestration layer we were building would be so crucial for the AI application. We now apply agents + sub agents and rules simultaneously to a

@levie swarms multiply noise as much as they multiply leverage. parallelizing slop doesn’t create clarity. it just accelerates confusion. most companies will drown in orchestration before they see real output. the future isn’t more agents, it’s deleting noise at the source. clarity

@levie This is exactly where that change management becomes critical.

@levie Absolutely, The potential of AI subagents to parallelize tasks is massive. Startups that build workflows AI-first are going to move at lightning speed, and even legacy companies that adapt will see huge gains. The future of work is looking extremely collaborative between humans

@levie agent coordination failures kill the leverage though spent months debugging context bleeding between agents in complex workflows. when one agent misunderstands another's output, the whole pipeline breaks single intelligent agent with comprehensive tool access usually

@levie Most big companies are not ready for that change yet, except for tech companies probably. The time to disrupt the stale industries is now.

@levie Totally agree. The paradigm shift is going from instructing a single assistant to orchestrating a team of specialists. It feels like the difference between writing a script and conducting an orchestra. What's the most surprising non-coding use case you've seen so far?

@levie I have only seen the angry teams messages after we receive the cloud bill

@levie You've nailed it—it's all about leverage. As a student trying to build these future systems, that leverage in learning is everything. A gifted Claude Pro or Chatgpt plus for a year would be transformative for my education, but it's currently out of reach for me.

@levie Swarm working on a very detailed simple task - yes of course Swarm working on a bit more complex task - hallelujahcination

@levie Agreed, transitioning to AI-first workflows will be a game-changer. Though, yeah, social media platforms might need to catch up first.

@levie Spot on! Subagent swarms feel like the next big unlock. Coding is just the start; once enterprises adapt workflows for AI-first, every corner of knowledge work gets transformed.