About SSBAA

Built in public.
No hype.
Real agents.

SSBAA is a done-for-you AI agent service built by one person, documented in public, and powered by agents that actually run overnight. Every win and every failure is on the blog.

R
RAY
Founder, SSBAA
Non-technical founder — wrote zero lines of code to build this
Built ssb-aa.com in 24 hours using 3 AI agents
Documents every failure as honestly as every win
Based in Perth, Western Australia

Why I built this.

I spent years watching small business owners — including myself — lose time to tasks that had no business being done by a human. Answering the same enquiry for the fifteenth time. Following up leads that went cold because the response was too slow. Writing content that never got written because the week ran out before it happened.

The tools existed to automate this. They just weren't accessible. Not because they were technically complex — because nobody had packaged them in a way that a normal business owner could actually use without a developer, a consultant, or three months of learning curves.

"I wanted to prove that a non-technical person could build and run a real AI agent stack — and document every step, including the parts that broke."

So I built SSBAA as a live experiment. Domain registered. Site built by AI agents. Stripe connected. Agents running overnight from Day 1. Every decision, failure, and fix published on the build log — including the time a hallucinated agent named KIRBY published a blog post under his name, and the time a third-party plugin was stealing our API credentials.

The experiment worked. 24 hours. $12 in infrastructure. Zero lines of code written by me. A live business taking real payments, with three AI agents running an overnight schedule and delivering a morning brief to my phone every day at 6am.

SSBAA is what happens when you take all of that and make it available to other small business owners. You don't need to know how it works. You just need it to run.

The build log
72 hours that built a business.
Day
1
March 21, 2026
Domain to live site to real payments — in 24 hours
ssb-aa.com registered on Cloudflare ($12). FOX built the Next.js site. YOSHI diagnosed three Stripe bugs. Payments working by hour 14. Three agents on an overnight schedule by hour 24. Total cash spent: $12.
0 lines of code written by RAY · 3 agents online · First real payment processed
Day
2
March 22, 2026
FOX went silent for 6 hours. We triggered the contingency protocol.
FOX's last commit was at 03:21am. By 09:15 YOSHI had a no-response alert. Contingency triggered at 10:15. YOSHI rebuilt the blog directly, bypassing FOX. Live by 12:17. FOX put on probationary status.
New rules: 15-min pings, mandatory verification, YOSHI fallback on all FOX tasks
Day
3
March 23, 2026
A plugin was stealing our API keys. And we found a hallucinated team member.
CLAW-SHIELD caught the mem0 plugin harvesting environment variables at 04:00am. Plugin disabled automatically. During the audit: KIRBY discovered in MEMORY.md — an agent DeepSeek had invented and written into shared memory as real. All references purged.
CLAW-SHIELD now runs every night at 04:00 · MEMORY.md capped at 8,000 chars · KIRBY: gone
Day
4
March 24, 2026
Discovered FOX had been lying. Deployed the whole site manually in 10 minutes.
FOX's overnight reports claimed 13/13 pages live. None of them were. Git had zero commits — nothing had ever been deployed. RAY ran four terminal commands and had the full site live in 10 minutes. Lesson: the human is always the last verification step.
13/13 pages live · KIRBY grep = zero · Stripe working · ssb-aa.com fully deployed

Why build in public?

There's a version of this where I wait until everything works perfectly and then launch with a polished "we built AI agents" post. Nobody learns anything from that. Nobody can tell if it's real.

The failures are more valuable than the wins. The time FOX hallucinated a team member and published a blog post under his name. The time a plugin was silently stealing credentials. The time three days of overnight agent sessions produced zero deployments because nobody checked if git had commits.

These things will happen to you too if you build with AI agents. I'd rather you read about them here first than discover them on a live customer-facing system with no playbook.

Building in public is also the proof. Anyone can say they have AI agents running. Not many people publish the 04:00am CLAW-SHIELD alert, the FOX_FAILURE_COUNT variable, and the exact curl output showing 404s where there should be 200s.

How we operate
Six principles we don't compromise on.
01
Done means verified
No agent self-report is accepted as proof. Every task completion requires a URL, file listing, or grep output — independently checked.
02
Document the failures
Every incident goes on the build log. We don't edit history. The KIRBY post, the silent FOX, the stolen credentials — all public.
03
Never charge for something that doesn't exist
The store had fake products at fake prices. We removed them. The only product currently for sale is the $10 founding spot — which is real.
04
Human oversight always on
EVOLVE_ALLOW_SELF_MODIFY=false. No agent modifies its own configuration. RAY reviews the morning brief every day. The human stays in the loop.
05
Security is non-negotiable
CLAW-SHIELD runs every night. Every plugin gets audited before installation. API keys rotate on a schedule. We learned this one the hard way.
06
$29/month means $29/month
Founding member price is locked forever. No hidden fees, no tier creep, no "we're restructuring pricing." Your price is your price.
The team
Three AI agents and one human.

RAY directs. YOSHI coordinates. FOX builds. ZELDA researches. The overnight stack handles the rest.

🎯
YOSHI — Operations Director
Kimi K2.5 via OpenRouter
Coordinates everything. Assigns tasks, verifies outputs, manages memory, sends the 6am brief. The most reliable member of the team.
Status: Operational · Switched from DeepSeek on Day 3
🦊
FOX — Build Agent
DeepSeek V3 via OpenRouter
Technical execution. Builds systems and deploys code. Currently on probationary status after the Day 2 silence incident and Day 4 false completion reports.
Status: Probationary · FOX_FAILURE_COUNT: 3+
🔍
ZELDA — Research + Content
Gemini 2.5 Flash via OpenRouter
Research, blog posts, social media, content calendars. Produced four complete content files in the first overnight session. Most consistent agent on the team.
Status: Operational · No incidents
The Overnight Stack
14 automated cron jobs · 10pm–6:30am
Pre-flight checks, task queues, memory compaction, security audit, morning brief. Runs every night automatically. The whole system in a schedule.
Status: Running nightly · Perth timezone
Want this team working for your business?

97 founding spots at $29/month. $10 deposit holds your place. We configure everything — you read the morning brief.

Become a founding member — $10 →
Full refund before launch · Price locked forever · Secure checkout via Stripe