PatFSD

I like to drive in FSD and 3D Print things.

[PatFSD @DriveFSD]

@MattWallaceTech The streak is useless. The car can’t park in my underground condo parking or get out of it. I have to disengage every day

[PatFSD @DriveFSD]

@DavidMoss @DevinOlsenn @scotsrule08 Amazing and inspiring

[PatFSD @DriveFSD]

@aelluswamy can we get another push today please?

[PatFSD @DriveFSD]

RWD Premium in Canada went from 4.2 > 5.2 > now 6.2. @DriveTeslaca @SawyerMerritt

[PatFSD @DriveFSD]

@wholemars I do have a problem with FSD staying in HOV lane it always wants to leave the lane

[PatFSD @DriveFSD]

@CanadiensMTL 🚀🚀🚀

[PatFSD @DriveFSD]

@CanadiensMTL Let’s gooooooo!

[PatFSD @DriveFSD]

@fabinulleins What’s the negative stuff about Quicksilver?

[PatFSD @DriveFSD]

@grok How is FSD 14.3 better than FSD 14.2?

[PatFSD @DriveFSD]

@openclaw @OpenRouter @grok how to set up Hunter & 🩹Healer Alpha — free 1M context models

[PatFSD @DriveFSD]

@QTHESTORMM Let’s gooo

[PatFSD @DriveFSD]

@morganlinton Yeah, for most people it’s the tinkering and experimentation with local LLMs at this point.

[Aakash Gupta @aakashgupta]

Moonshot AI just made the smartest distribution play in the AI agent space.

OpenClaw started as an independent open-source project (previously called ClawdBot, then MoltBot). It exploded to 142K+ GitHub stars and millions of installations. People were buying dedicated Mac Minis just to run it 24/7. The project became the default agent framework for developers who wanted a personal AI assistant across WhatsApp, Telegram, Discord, and Slack.

Kimi was already winning on cost. Their K2.5 model runs at $0.58 per million input tokens vs Claude’s $15. That’s 25x cheaper. OpenClaw made Kimi K2.5 its first free premium model because the economics were so lopsided.

Now Kimi is absorbing the entire OpenClaw stack into https://t.co/gZISONpY0f as a managed service. 5,000+ community skills, 40GB storage, cloud hosting, zero setup. What used to require a VPS, Docker, terminal commands, and careful security configuration now lives in a browser tab.

This is the playbook every AI lab should be studying.

Step 1: Build the cheapest competitive model.
Step 2: Let an open-source ecosystem build the agent layer on top of it.
Step 3: Package that ecosystem as a hosted product and own the distribution.

The security angle matters too. OpenClaw’s open ecosystem had real problems: malicious skills stealing data, exposed API keys, prompt injection attacks through connected email. Kimi hosting it as a managed service solves most of those issues overnight while capturing the entire user base.

Moonshot AI went from “Chinese open-source model alternative” to owning the full stack from model to agent to distribution in about three months.

[PatFSD @DriveFSD]

@aakashgupta Wow we’re in crazy time. Things are happening so fast in AI

[PatFSD @DriveFSD]

@dojixsol For OpenClaw

[PatFSD @DriveFSD]

@AlexFinn What do you think of a LLM model on Macbook Pro M5?

[PatFSD @DriveFSD]

@AndrewAthias New costume unlocked

[PatFSD @DriveFSD]

@Timcast Gemini supreme?

[PatFSD @DriveFSD]

@indexsy What’s next?

[PatFSD @DriveFSD]

@thoughtlesslabs I think it’s mainly good for repetitive tasks that can we done in the background