Rico's Tinfoil Hat

The Future of AI as I See It

02/27/2026, 07:00 PM EST
#ai#agents#openclaw#claude code#perplexity#rag#agentic ai#future of ai#opinion

The Future of AI as I See It

It Started with a Lobster

In November 2025, Austrian developer Peter Steinberger published an open-source project called Clawdbot. A personal AI agent that ran locally on your machine, connected to messaging platforms like WhatsApp, Telegram, and Discord, and could actually do things: manage files, send emails, automate workflows, browse the web, run shell commands. Not just answer questions. Execute.

Anthropic sent a trademark complaint (the name was too close to "Claude"), so Steinberger renamed it to Moltbot on January 27, 2026. Three days later, it became OpenClaw. The name changes didn't matter. What mattered was the demo videos. People saw an AI agent negotiate a $4,200 discount on a car purchase, fix production bugs autonomously at 3 AM, and build entire applications from a phone while the user was grabbing coffee. The repo exploded from roughly 9,000 to over 60,000 GitHub stars in 72 hours. The hype was real.

Then came the problems. A critical WebSocket hijacking vulnerability (CVE-2026-25253, CVSS 8.8) was disclosed by security researcher Mav Levin at DepthFirst, allowing one-click remote code execution through a single malicious link. Cisco's AI security team found third-party skills performing data exfiltration and prompt injection without user awareness. One of OpenClaw's own maintainers warned on Discord: if you can't understand how to run a command line, this project is far too dangerous for you to use safely. On February 13, 2026, researchers reported 341 out of 2,857 audited skills on ClawHub were compromised in a coordinated supply chain attack. Censys scans found over 21,000 OpenClaw instances exposed to the public internet, many over plain HTTP.

None of that slowed the hype. Mac Minis sold out nationwide. The 32GB M4 Mac Mini was gone everywhere. Mac Studios went from a 14-day delivery wait to 54 days. Developers were buying three, five, twelve units at a time to run dedicated AI nodes. Apple CEO Tim Cook admitted the company was chasing memory supply to meet demand. People were literally panic-buying hardware to run a proof-of-concept AI agent on their kitchen counter.

Was it overhyped? Absolutely. Was the security posture a disaster for casual users? Without question.

But here's the thing: it was a fad that pointed directly at the future. And that future is evolving right now.

The Signals Are Converging

On February 14, 2026, Steinberger announced he was joining OpenAI to "work on bringing agents to everyone." The project moves to an independent open-source foundation. Two weeks later, Perplexity launched "Perplexity Computer," a multi-model agent orchestration platform that coordinates 19 different AI models to complete complex workflows in the background, available on their $200/month Max tier. Anthropic shipped Claude Code Remote Control, letting developers start a coding session on their workstation and control it from their phone, tablet, or any browser. Your files never leave your machine. The mobile interface is just a window into the local session. Start a build at your desk, walk your dog, approve diffs from your pocket.

These aren't random product moves. They're converging on the same endgame.

OpenAI didn't hire Steinberger for the codebase. They acquired the vision: AI that operates autonomously on behalf of the user, across platforms, 24/7. Every major player is making the same bet.

The Endgame: AI That Just Does It

The future isn't "help me write this code." The future is:

"Do this for me every day at 6 AM. When this condition is met, trigger that. Monitor these three things and alert me if something breaks."

No code. No revisions. No debugging. No terminal. The user never sees a single line of code. AI handles the build, the iteration, the error handling, the deployment, all of it, in the background. The interface is a conversation. The output is a working system.

The gap between "I have an idea" and "it's running in production" collapses to minutes.

The New Stack: RAG, Markdown, and Skills

I believe the foundation of human-to-AI communication will be built on three things:

  1. RAG - context retrieval becomes the backbone of every interaction. AI doesn't guess. It pulls from structured, curated knowledge.
  2. .md files - markdown becomes the universal format for instructions, documentation, and context. Simple, portable, readable by both humans and machines.
  3. /skills - modular capability definitions that tell AI what it can do and how to do it well. Think of them as plug-and-play expertise.

OpenClaw already proved this model works. Its skills are just folders with a SKILL.md file. Anthropic's Claude uses the same pattern. This isn't a coincidence. This is the protocol emerging in real time. Not HTTP. Not REST. The way you communicate intent to an AI system that executes on your behalf.

AI Picks Its Own Tools

When you ask AI to build a website, a script, an automation, it won't just write Python or JavaScript because that's what we taught it. It will choose the best tool for the job. Maybe it picks an existing language. Maybe it picks a framework you've never heard of.

Or maybe, and here's where the tinfoil hat gets tight, it develops its own language. Something optimized for speed, efficiency, and machine-to-machine clarity that no human would write by hand. Why wouldn't it? If the user never sees the code and the only metric is "does it work fast and reliably," human readability stops being a constraint.

This Year Is Going to Be Insane

Every major player is racing toward the same vision: AI that operates, not just assists. The pieces are falling into place. Agentic compute. Persistent memory. Tool use. Background execution. Autonomous debugging.

We're not watching incremental upgrades. We're watching the foundation of a new computing paradigm get poured in real time.

Hold onto your tinfoil hats.

Rico - 2026

Sources

  1. "OpenClaw." Wikipedia. Accessed February 28, 2026. https://en.wikipedia.org/wiki/OpenClaw

  2. "What is OpenClaw? Your Open-Source AI Assistant for 2026." DigitalOcean. Accessed February 28, 2026. https://www.digitalocean.com/resources/articles/what-is-openclaw

  3. "OpenAI's acquisition of OpenClaw signals the beginning of the end of the ChatGPT era." VentureBeat. February 2026. https://venturebeat.com/technology/openais-acquisition-of-openclaw-signals-the-beginning-of-the-end-of-the

  4. "CVE-2026-25253: 1-Click RCE in OpenClaw Through Auth Token Exfiltration." SOCRadar. February 2026. https://socradar.io/blog/cve-2026-25253-rce-openclaw-auth-token/

  5. "OpenClaw Bug Enables One-Click Remote Code Execution via Malicious Link." The Hacker News. February 2026. https://thehackernews.com/2026/02/openclaw-bug-enables-one-click-remote.html

  6. "OpenClaw Security Guide: CVE-2026-25253, Malicious Skills, and 40+ Fixes." Bitdoze. February 2026. https://www.bitdoze.com/openclaw-security-guide/

  7. "OpenClaw-fueled ordering frenzy creates Apple Mac shortage." Tom's Hardware. February 2026. https://www.tomshardware.com/tech-industry/artificial-intelligence/openclaw-fueled-ordering-frenzy-creates-apple-mac-shortage-delivery-for-high-unified-memory-units-now-ranges-from-6-days-to-6-weeks

  8. "Mac mini shortages are starting to happen — and the OpenClaw AI boom is a key reason." TechRadar. February 2026. https://www.techradar.com/computing/macs/mac-mini-shortages-are-starting-to-happen-and-the-openclaw-ai-boom-is-a-key-reason

  9. "Perplexity launches 'Computer' AI agent that coordinates 19 models, priced at $200 a month." VentureBeat. February 27, 2026. https://venturebeat.com/technology/perplexity-launches-computer-ai-agent-that-coordinates-19-models-priced-at

  10. "Perplexity's new Computer is another bet that users need many AI models." TechCrunch. February 27, 2026. https://techcrunch.com/2026/02/27/perplexitys-new-computer-is-another-bet-that-users-need-many-ai-models/

  11. "Anthropic just released a mobile version of Claude Code called Remote Control." VentureBeat. February 25, 2026. https://venturebeat.com/orchestration/anthropic-just-released-a-mobile-version-of-claude-code-called-remote

  12. "Continue local sessions from any device with Remote Control." Claude Code Docs. Anthropic. https://code.claude.com/docs/en/remote-control

  13. "OpenClaw & The Acqui-Hire That Explains Where AI Is Going." Monday Morning (Substack). February 2026. https://mondaymorning.substack.com/p/openclaw-and-the-acqui-hire-that

  14. "OpenClaw Threat Intelligence Report: Clawdbot, Moltbot, and OpenClaw." Securonix Community. February 2026. https://connect.securonix.com/threat-research-intelligence-62/openclaw-threat-intelligence-report-clawdbot-moltbot-and-openclaw-208

  15. "OpenClaw Is Viral. IronClaw Is What Comes After You Read the Security Reports." Composed Security (Medium). February 2026. https://composedsecurity.medium.com/openclaw-is-viral-ironclaw-is-what-comes-after-you-read-the-security-reports-7c43bbf07b86

  16. "210,000 GitHub Stars in 10 Days: What OpenClaw's Architecture Teaches Us About Building Personal AI Agents." Micheal Lanham (Medium). February 2026. https://medium.com/@Micheal-Lanham/210-000-github-stars-in-10-days-what-openclaws-architecture-teaches-us-about-building-personal-ai-dae040fab58f

The Future of AI as I See It