Alan VanToai

Serial founder (3x Exits). Supporting enterprise leaders with AI transformation @ AI Performance Lab.

[The Humanoid Hub @TheHumanoidHub]

Figure 03 comes with a specialized charging stand.

It anchors the robot so the motors can relax and stop using power to balance, while charging it inductively at 2 kW through the soles of its feet.

Robots can do 4 to 5 hours of work after 1 hour of charging.

[Brett Adcock @adcock_brett]

Robot Fashion

[Alan VanToai @VanToai]

I spoke with a company exec last weekend who said something I keep hearing:

"We just got set up with Claude. What happens if Codex ships something better next week? Or Gemini in three months? How do we keep up?"

This fear is rational, and also slightly misdirected.

You're not locked into a model. You're locked into the prompts, tools, memory, and workflows you've wrapped around the model. Which means the answer isn't "pick the right model" — it's "build the wrapper so the model becomes the swappable part."

That wrapper has a name. It's a harness.

A modern agent harness has six pieces:

1. Memory — durable storage for what the agent has learned across sessions. User preferences, project context, prior decisions. Lives in your database, not in the model's context window.

2. Context files — the system prompts, role definitions, and retrieved knowledge the agent gets at runtime. Versioned in your repo, not hardcoded into one provider's playground.

3. Skill files — reusable workflows and procedures the agent can call. "How we onboard a customer." "How we triage a bug." These are durable assets that outlive any model.

4. Tool connections — MCP servers and direct API integrations to your systems of record. Slack, GitHub, your CRM, your data warehouse. Standardized so any model can call them.

5. Model gateway — a thin routing layer (LiteLLM, OpenRouter, or a fifty-line proxy) that normalizes auth, logging, and fallback across providers. This is the load-bearing piece that makes "swappable" actually work.

6. Eval harness — twenty to fifty golden test cases per agent that you run on every prompt change and every model swap. This is the thing that turns "switch to the new Claude" from a scary launch into a Tuesday.

That's the harness. Everything in it is yours. Everything in it is durable.

The model is the engine that drops into the chassis.

When Anthropic ships the next thing, you don't rebuild your agent. You change a config line, run your evals, and ship. When Gemini leapfrogs in six months, same move. The harness doesn't care. That's the whole point.

The teams that feel locked in usually skipped the harness. They're calling the model API directly from their app, with prompts hardcoded in production code, no eval suite, no gateway. Of course they feel locked in — they didn't build the layer that makes models interchangeable.

The principle:

Models will change faster than your business processes. Build the layer that's durable. The model is supposed to be the swappable part.

If you're worried about being leapfrogged, that worry is the signal that you've been investing in the wrong layer. The fix isn't to bet harder on a single provider. The fix is to build the harness — and let the model be a config change, not a strategic decision.

The companies that win the next few years won't be the ones who picked the right model. They'll be the ones who built the right harness around whatever model is best this quarter.

[Alan VanToai @VanToai]

@vasuman Great post, Vas. Very sobering & inspiring.

Sent you a DM.

[Alan VanToai @VanToai]

@perplexity_ai One of the slickest onboardings I've ever experienced. Thank you. Great work!

Bug report: my Computer task history isn't populating in the Mac app. Neither Computer tasks from before, nor tasks originated in the Mac app itself.

Perplexity Questions are populating fine.

[Alan VanToai @VanToai]

@yazins Would love to interview you about this! Check DMs :)

[Alan VanToai @VanToai]

@_saberamani Thanks for the question Tim. Yeah man, it's a repeatable process. I'm building this into a repeatable skill that I can execute over and over again for new clients.

[Alan VanToai @VanToai]

@Accelr8_Dan It’s an incredibly powerful unlock and worth the effort.

My two hacks:

1. Find other examples and see what other people are doing. I’ll share mine with you!

2. Yes, tinker until you like the output. Your discernment and taste is the final boss.

[Alan VanToai @VanToai]

@Gavriel_Cohen @CNBC Very well said, @Gavriel_Cohen. Appreciate the values you’re building with here. Keep up the great work.

[Alan VanToai @VanToai]

Loving the thinking and framing here by @karpathy.

Source code configurability by self-improving Claws. Claws as a new layer on top of LLMs & agents. Physical device “possessed by the soul of a personal digital house elf” ☺️

Delightful.

[Andrej Karpathy @karpathy]

Bought a new Mac mini to properly tinker with claws over the weekend. The apple store person told me they are selling like hotcakes and everyone is confused :)

I'm definitely a bit sus'd to run OpenClaw specifically - giving my private data/keys to 400K lines of vibe coded monster that is being actively attacked at scale is not very appealing at all. Already seeing reports of exposed instances, RCE vulnerabilities, supply chain poisoning, malicious or compromised skills in the registry, it feels like a complete wild west and a security nightmare. But I do love the concept and I think that just like LLM agents were a new layer on top of LLMs, Claws are now a new layer on top of LLM agents, taking the orchestration, scheduling, context, tool calls and a kind of persistence to a next level.

Looking around, and given that the high level idea is clear, there are a lot of smaller Claws starting to pop out. For example, on a quick skim NanoClaw looks really interesting in that the core engine is ~4000 lines of code (fits into both my head and that of AI agents, so it feels manageable, auditable, flexible, etc.) and runs everything in containers by default. I also love their approach to configurability - it's not done via config files it's done via skills! For example, /add-telegram instructs your AI agent how to modify the actual code to integrate Telegram. I haven't come across this yet and it slightly blew my mind earlier today as a new, AI-enabled approach to preventing config mess and if-then-else monsters. Basically - the implied new meta is to write the most maximally forkable repo and then have skills that fork it into any desired more exotic configuration. Very cool.

Anyway there are many others - e.g. nanobot, zeroclaw, ironclaw, picoclaw (lol @ prefixes). There are also cloud-hosted alternatives but tbh I don't love these because it feels much harder to tinker with. In particular, local setup allows easy connection to home automation gadgets on the local network. And I don't know, there is something aesthetically pleasing about there being a physical device 'possessed' by a little ghost of a personal digital house elf.

Not 100% sure what my setup ends up looking like just yet but Claws are an awesome, exciting new layer of the AI stack.

[Alan VanToai @VanToai]

Incredible work here. Nice one, @airtable

[Howie Liu @howietl]

https://t.co/IZJWnHj5UW

[Alan VanToai @VanToai]

Sheeeesy boys be out here cookin 🔥🔥 nice one @howietl 💪🏾💪🏾

[Howie Liu @howietl]

I've been personally burning through billions of tokens a week for the past few months as a builder. Today I'm excited to announce Hyperagent, by Airtable.

An agents platform where every session gets its own isolated, full computing environment in the cloud — no Mac Mini required. Real browser, code execution, image/video generation, data warehouse access, hundreds of integrations, and the ability to learn any new API as a skill.

Deep domain expertise through skill learning. Teach the agent how your firm evaluates startups or how your team runs due diligence — now anyone on the team gets output that reflects your actual methodology, not a generic template.

One-click deployment into Slack as intelligent coworkers. These aren't bots that wait to be @mentioned — they follow conversations, understand context, and act when relevant.

And a command center to oversee and continuously improve your entire fleet of agents at scale.

We're onboarding early users now. https://t.co/kctMfFCQqG

[Alan VanToai @VanToai]

Easily my favorite piece on anything OpenClaw/agent related.

The role that examining & re-patterning my core beliefs and stories plays in my own growth journey makes this hit insanely hard.

Souls >
Skills

[toli @tolibear_]

https://t.co/gprOAK0k9n

[Alan VanToai @VanToai]

@tolibear_ Easily my favorite piece on anything OpenClaw/agent related.

The role that examining & re-patterning my core beliefs and stories plays in my own growth journey makes this hit insanely hard. It’s what I am doing with @lucidvision_app

Very much inspired here. DMing you.

[Aakash Gupta @aakashgupta]

Bostrom and Kurzweil mapped this out decades ago. There are roughly three phases between here and abundance, and your financial strategy should look completely different in each one.

Phase 1 is now through ~2030. AI eats white-collar tasks. Costs drop in specific sectors. Scarcity persists in housing, energy, healthcare, and physical goods. Traditional saving still works here because the economy still runs on the old rails.

Phase 2 is the messy middle. Maybe 2030 to 2038. Humanoid robots start scaling. Autonomous systems handle logistics, manufacturing, construction. Costs plummet across categories. This is where Musk’s thesis starts getting interesting. Your 401(k) contributions are buying assets denominated in a currency whose purchasing power is shifting underneath you. Saving “money” might matter less than owning productive assets, energy capacity, or compute.

Phase 3 is full post-scarcity. Production approaches zero marginal cost. Kurzweil’s law of accelerating returns hits escape velocity. If this arrives, the thread is right. Your brokerage account is a relic.

Nobody knows the transition speed between phases. The jump from Phase 1 to Phase 3 could take 5 years. Could take 50. Vernor Vinge called this the “event horizon” problem. You cannot see past the singularity because the rules change too fast to model.

The smartest play: save aggressively in Phase 1 (you’re in it), but tilt your portfolio toward assets that appreciate during the transition. Companies building the infrastructure of abundance. Energy. Compute. Robotics. Physical AI.

You’re betting on surviving the lag between when old systems break and new ones arrive.

The sci-fi version: the starship is coming, but you still need oxygen for the walk to the launch pad.

[Alan VanToai @VanToai]

This excites me and haunts me:

Today, literally anything and everything is just a prompt (or series of prompts) away.

@steipete prompted @openclaw to life & changed the world.

Each of us has the power to do the same.

It’s just a matter of knowing which questions to ask.

[Alan VanToai @VanToai]

.@MatthewBerman absolutely cooked 🔥

[Matthew Berman @MatthewBerman]

I'm one of the most advanced users of OpenClaw.

OpenClaw + GPT5.3 Codex + Opus 4.6 has been the trifecta that changed everything.

I made a video going over everything I'm doing with these tools.

Learn these tools, stay ahead.

Watch this video right now.

0:00 Intro
1:02 Overview
4:17 Sponsor
5:12 Personal CRM
7:11 Knowledge Base
8:30 Video Idea Pipeline
11:09 Twitter/X Search
12:47 Analytics Tracker
13:33 Data Review
15:34 HubSpot
16:13 Humanizer
16:52 Image/Video Generation
18:22 To-Do List
19:37 Usage Tracker (Saves Money)
20:45 Services
21:25 Automations
22:42 Backup
23:30 Memory
24:06 Building OpenClaw
25:22 Updating Files

[Alan VanToai @VanToai]

@jforjacob Skill issue…

[Alan VanToai @VanToai]

@TechWith_Nova V2

[Alan VanToai @VanToai]

@bc1beat Love yours!