Denis

Building @FlyMy_AI - everything your agent needs in one cloud.
All models, MCPs & serverless agents. One MCP.
(っ˘з(˘⌣˘ ) (⊙_⊙) ᓚ₍ ^. .^₎ ⊂(・▽・⊂)

ex @nvidiaai

[Denis @cop_on_fire]

Introducing MCP Sharing, built into FlyMyAI infra.

Configure FlyMy MCP once: 1500+ MCPs inside, from Telegram and Slack to Instagram and Brex.
Use it from Claude, Codex or any agent.

Share it with your team.

Salesforce seat: $350/user/mo.
Shared MCP: from $0.001 per call.

> work on one Notion page, one Linear board, one CRM with the whole team
> check and answer email from one inbox
> give a contractor 7 days - access ends on its own
> share access, not keys

claude "connect using https://t.co/oZJwRqpqmv"
claude "share my notion with [email protected] for 7 days"

One MCP. Everything your agents need. No seats.

[Denis @cop_on_fire]

@nrx838 Follow for more use-cases

[Denis @cop_on_fire]

Hundreds of new AI models drop every week. Most of them are noise.

We pointed an agent at it and told it to surface the ones that actually matter across LLM, image, video, audio and 3D. Then we froze the run so it wouldn’t just die in the chat.

Now it has an API and a schedule. Close the laptop, it still runs.

[Denis @cop_on_fire]

https://t.co/5HqT4dprB0

[Denis @cop_on_fire]

@Tiny_Fish @lightsagehq so the README is the landing page now?

[Denis @cop_on_fire]

@venturetwins @GoogleDeepMind trash a room, leave, then come back. Does it remember the mess?

[Denis @cop_on_fire]

4/ Play them yourself. Code and prompts inside, clone any and make your own:

Foldable iPhone
Opus: https://t.co/IuqHgaxSpC
Kimi: https://t.co/PkD0a0BBOt

Jelly Minecraft
Opus: https://t.co/stb9dtc4BQ
Kimi: https://t.co/6c9nlOZd4r

ASCII Dark Souls
Opus: https://t.co/ceuOQwig7z
Kimi: https://t.co/ERulxm78mY

Same starting prompt, different coding clients (Opus in Claude Code, Kimi via OpenRouter). Not a controlled benchmark. Kimi's Jellycraft includes one follow-up fix. Footage at 1x, ASCII clips are built-in showcase modes.

[Denis @cop_on_fire]

On paper Kimi K3.1 Pro matches Opus 5.5 at a third of the cost.

In my builds it's not even close.

Their Terminal-Bench run used Claude Code's harness. I gave both the same prompt for three browser games:

1/
Foldable iPhone: Opus closes into a phone with a lock screen. Kimi closes into a grey slab.

[Kimi.ai @Kimi_Moonshot]

Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.

🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.

Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on https://t.co/85jA71X9gw, Kimi Work, Kimi Code, and the Kimi API.

Open weights by Nov 4, 2026.

🔗 API: https://t.co/XCrgjXAqMw
🔗 Tech blog: https://t.co/5iPP0MBej6

[Denis @cop_on_fire]

3/
ASCII Dark Souls: Opus draws a boss with teeth. Kimi's boss you have to squint at

[Denis @cop_on_fire]

@boardyai Sweet :) thx @boardyai

[Denis @cop_on_fire]

Production agents don't want another dashboard.

They need somewhere to keep working:
→ memory that survives the chat
→ tools with permissions
→ retries and schedules
→ an API your product can call

The model is rented. The execution body is the product.

[Denis @cop_on_fire]

@EllaTech0 💯

[Denis @cop_on_fire]

3 points behind Opus 5.5. $2 / $10 per 1M tokens.
Open weights by Nov 4.

That's Kimi K3.1 Pro, the same week Anthropic heads into a ~$2T IPO.

The model is rented.
Make sure you can switch landlords.

[Kimi.ai @Kimi_Moonshot]

Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.

🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.

Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on https://t.co/85jA71X9gw, Kimi Work, Kimi Code, and the Kimi API.

Open weights by Nov 4, 2026.

🔗 API: https://t.co/XCrgjXAqMw
🔗 Tech blog: https://t.co/5iPP0MBej6

[Denis @cop_on_fire]

@ZorayaAi Yep

[Denis @cop_on_fire]

@aaliya_va Yep. Open to chat about your agentic use-cases @Aaliya

[Denis @cop_on_fire]

@Pkhromov5511 Thanks Pavel! Exactly, without memory agents are still blind in the long run.

[Denis @cop_on_fire]

@nrx838 Ofc 🚀

[Denis @cop_on_fire]

@burninganna Congrats guys!

[Denis @cop_on_fire]

@Pkhromov5511 @FlyMy_AI Looks cool! Thanks for mentioning us!
Share more! We've added credits on your account 🙃

[Denis @cop_on_fire]

We built Artifacts into @FlyMy_AI.

Think Claude Artifacts on steroids.

Here’s what you can do with them:
→ run an agent, a model, or a typed decision right from the page
→ build games, slides, landing pages, and more
→ keep API keys out of the page
→ publish new versions while keeping older ones listed
→ keep sharing off until you turn it on
→ invite specific people. Agents can edit too
→ turn remixing on or off
→ use HTML, JS, CSS, Three.js, images, audio, and WebAssembly

What you can ship:
→ an internal tool your team runs from a link
→ an MCP demo people can try without touching a config
→ a report your agent rebuilds, with every version kept
→ a form where an agent scores every answer

What should your agent ship as a page?