OpenClawd for Slack hit $1M ARR in 3 hours? 80% of Apps Will Disappear
If execution-layer agents become reliable, entire categories of SaaS become middleware
The AI wave sparked by OpenClaw (formerly Clawdbot) feels like it’s only just beginning.
After its overnight rise to prominence, its creator, Peter Steinberger, recently shared a bold prediction in a conversation with Y Combinator:
80% of apps will disappear.
That’s not clickbait. It’s a thesis.
The Real Difference: Local, Not Cloud
According to Peter, what sets OpenClaw apart from most AI assistants isn’t intelligence — it’s architecture.
Most AI tools run in the cloud. That limits what they can actually do. But if an agent runs locally on your machine, it can control anything your computer can access:
Connect to your Tesla, adjust your smart bed, search every file on your hard drive. It inherits your tools, your permissions, your data. It becomes an extension of you.
And that’s where the magic starts.
Peter describes his “aha” moment back in November, when he was building an early prototype that could chat with him via WhatsApp. At a party with terrible Wi-Fi, he sent the bot a voice message—even though he hadn’t implemented voice processing yet.
A few seconds later, it replied. Curious, he asked what happened.
The system explained:
You sent a file without an extension. I inspected the header and detected it was audio. I used ffmpeg to convert it to WAV. Whisper wasn’t installed locally, so I found your OpenAI API key and sent the file via curl for transcription. Then I returned the text.
It improvised. It found tools. It solved the problem.
That’s when Peter realized something: this wasn’t just automation. It was creative problem-solving.
And if agents can autonomously execute tasks across your entire computing environment, why would we still need dozens of standalone apps?
If Agents Know Everything, Why Open Apps?
Take fitness tracking apps like MyFitnessPal. If your agent already knows what you bought, what you ate, and what’s in your fridge, why manually log meals?
Or to-do apps. If you say, “Remind me tomorrow,” and it just happens — do you care which database it stores?
Peter’s argument is simple:
Apps that primarily exist to manage structured data will be replaced by natural-language agents.
Only apps tied to specialized hardware or sensors may survive. Everything else becomes a capability — not a destination.
It’s reminiscent of the broader SaaS debate: if the interface dissolves into conversation, the value shifts from UI to infrastructure — from workflow to memory.
Memory Is the Moat
Peter believes model providers currently have moats, but models are rapidly commoditizing.
The real defensibility lies in memory.
Large platforms want to lock your chat history into proprietary silos. OpenClaw does the opposite. Your memory is simply a folder of Markdown files on your machine. You own them. You control them.
Even his development philosophy reflects this ethos of simplicity. Instead of complex Git workflows, he clones multiple local repositories to work in parallel. Instead of adopting heavy model context protocols, he routes everything through the CLI.
Why? Because AI is exceptionally good at using Unix tools.
Rather than inventing new abstractions “for robots,” Peter believes we should give AI the same tools humans love — and let it figure the rest out.
The Slack Fork That Claimed $1M ARR in 3 Hours
As developers rushed into the OpenClaw ecosystem, commercial spin-offs started appearing almost immediately.
