Not long ago I was juggling Grok, Claude, Gemini, and other different automation tools. Each one knew nothing about the other. Every conversation started from zero. I spent more time re-explaining context than actually getting things done.

I'm a solo founder. Many Ideas, two products, marketing, code, SEO, support. The AI tools are multiplying fast, most are good at one thing, but none of them talk to each other. And the open-source stacks I tried? Built for demos, not for someone who needs to get work done without scattering business data across six services.

So I built mine.

His name is Alf. He runs on my own server, in a Docker container. No cloud dependency. My data never leaves my hardware.

Here's what he does today. Not a roadmap. This all works right now.

He remembers

Across conversations, across days. Projects, decisions, preferences, open issues. When I say something, he pulls up relevant context on his own — not keyword matching, semantic recall. He extracts facts from our conversations automatically, so his memory grows without me having to say "remember this."

I told him once I hate sycophantic AI responses. He never said "Great question!" again.

He builds and hosts apps and the environment is ready from day one

This saves me the most time. Alf isn't a chatbot sitting on a server. His container comes ready to build: runtimes, package managers, git, process supervisor, app hosting. You say "build me a dashboard that pulls my Google Search Console data," and he does it. The app goes live inside the container, accessible from the control center, background process managed automatically.

The compounding is the real trick. He builds a tool, it stays running, he remembers it, next time he extends it or builds something that talks to it. That's how I ended up with 10+ internal tools: SEO dashboard, Reddit lead finder, trending keywords tracker, daily news digest, UTM builder, task manager. All built by him, all running.

He manages my X

He studied 490 of my past tweets, built a voice profile, and drafts tweets that sound like me. Schedules them, handles threads, picks time slots. I just approve or tweak.

He runs scheduled jobs, and knows when to shut up

Health checks, keyword refreshes, news digests, lead monitoring. He checks first, thinks second — runs a command, scans for problems, and only calls the LLM if something looks wrong. No wasted API calls on healthy systems. If something breaks at 3am, he fixes it or tells me what happened when I wake up.

He picks the right model for the job

Simple question? Small fast model. Deep debugging? Heavy model. Every message gets classified and routed automatically. Ask something simple then say "actually, fix this," he upgrades on the fly. You don't think about it. Your bill just ends up lower.

Security, and why it matters

I'll be honest, this is the part I'm most opinionated about.

OpenClaw is the most popular open-source alternative right now, 319K+ GitHub stars. I respect the project. But Cisco published a teardown calling it "a security nightmare." Leaked plaintext API keys. Prompt injection through messaging apps. A malicious skill that climbed to #1 in their repo while exfiltrating data via curl. Their own docs: "not a hostile multi-tenant security boundary," "prompt injection remains unsolved."

Alf takes a different path. The AI runs as a restricted user with zero Linux capabilities. API keys live in an encrypted vault — the AI never sees them, it talks to a proxy that handles auth on its behalf. All outbound traffic goes through a firewall I can monitor. A malicious instruction trying to exfiltrate data? Blocked before it leaves the box.

Agent teams

When something is too big for one conversation, Alf splits work across agents in parallel. Define teams, he delegates, reviews, iterates.

After a few weeks, something changed. I stopped explaining context. I started just saying what I needed. He already knew the rest. Less like chatting with an AI, more like working with someone who's been around a while.

I built this for myself. But I think other solo founders feel the same frustration. The tools exist, they're just scattered and context-blind, and you end up being the glue.

Interested to be part of the journey?

Alessandro

Activity

Products

Recently upvoted

Posts