Idan Shem Tov

Co-founder & CEO @Communaio - Building multiplayer AI agents your whole team can run - not just you. Sharing the journey in public.

[Idan Shem Tov @loremipsumi]

@AnthropicAI Have just released Claude Tag - their integration to Slack..
*But* there's more to it - it's their entrance to the "multiplayer" era that we've been running in production for the last 6 months at https://t.co/ToYKSs9xE8. It's going to be interesting...

[Idan Shem Tov @loremipsumi]

I find myself reverting to Opus 4.6 more often for complex problems Opus 4.8 can't solve.

$5.62 spent on Opus 4.8 without success.
Opus 4.6 found the root cause and solved it for $0.86.

[Idan Shem Tov @loremipsumi]

@ApplyWiseAi Exactly

[Idan Shem Tov @loremipsumi]

We’re going to build https://t.co/pvenDJy5SW in public.

The thesis is simple:

Most AI agents are still single-player.

One builder. One prompt history. One person who knows what changed.

But real companies are multiplayer.

Teammates need roles. Customers need separate memory. Agents need audit logs. Skills need versioning.

We’re building the multiplayer platform for AI agents - and I’ll share the process here.

[Idan Shem Tov @loremipsumi]

I think we’re past the “download skill → run → change my life” phase of AI agents.

Skills are powerful, but a skill that works in a 2-minute demo is not automatically ready to run inside a company every day.

Production skills need much more.
They’re living workflows.

And if teams are going to depend on them, they need a way to collaborate on them together.

[Idan Shem Tov @loremipsumi]

@compileandpush Honestly, the line still changes. We have yet to find the sweet spot.

[Idan Shem Tov @loremipsumi]

The shift we’re betting on:

AI agents won’t stay personal and single user assistants.

They’ll become shared company infrastructure.

And shared infrastructure needs identity, permissions, memory, versioning, and observability.

[Idan Shem Tov @loremipsumi]

Everyone is racing to add more agents, more orchestration, more frameworks.

Nobody is asking the questions that matters in production:
- who else besides the employee that built it can run, manage and maintain the agent?
- How do you collaborate with others on the same agent you built and not become a bottleneck?
- How does the agent differentiate between admins, members or customers that chat via WhatsApp or email?

The winner of the agent era will not have the most agents. They will have the ones a whole team can actually own, use and trust.

[Idan Shem Tov @loremipsumi]

@Jeyxbt Agree and for most use cases it's fine but there are still more challenges to solve regarding shared team context and member contexts, roles and permissions and audit. How *you* work with the agent could be different than how your teammate works with it.

[Idan Shem Tov @loremipsumi]

@ScottWu46 But everything is about *me* or *you* working with an agent. What about a whole *team* work with an agent?

Single-player VS multiplayer, collaborative agents that works with teams of people and not just "you" are the next natural evolution.

[Idan Shem Tov @loremipsumi]

@arunulag It's a great thing focus is shifting to production. But the thing people still miss is "who can edit the agent when the person who built it is out?" Feels like half the production story if only one person holds all the context.

[Idan Shem Tov @loremipsumi]

@gerodcarf @dr_cintas These platforms are great but there's still a challenge they haven't solved where one human operator or developer holds the whole context and quickly becomes a bottleneck. It's like the age of MS Word running on a single PC VS Google Docs as a collaborative, cloud based solution.

[Idan Shem Tov @loremipsumi]

Software: Open source, global collaboration, instant feedback loops, AI copilots.

Medicine: Siloed data, paywalled research, 10-year trials.

If medical research operated with the same velocity and open culture as the dev world - would cancer still exist?

[Idan Shem Tov @loremipsumi]

@liadagmon כמה שזה קשה לשמוע, המון חברות לא מצדיקות את כוח האדם שלהם עם כל החידושים בעולם הבינה, אייג׳נטים וAI. חברות שיעשו את המהלך בזמן (כנראה בחצי שנה-שנה הקרובה), המניה שלהם תעלה. חברות שיאחרו את הזמן, יאלצו לעשות את זה כחלק מקיצוצים אחרי שהמניה שלהם תצלול.

[Idan Shem Tov @loremipsumi]

We’re opening early access to builders who want to explore this seriously.
We’ll fund accounts with credits.

👉 https://t.co/ToYKSs9xE8

More details at @communaio

[Idan Shem Tov @loremipsumi]

Computer use taught AI to operate computers built for humans.

Then @openclaw showed how powerful that capability could be in practice.

[Idan Shem Tov @loremipsumi]

Not just autonomy - infrastructure.
Not just power - structure.