48 hours after I announced a £2 million investment proposal to shareholders and my team, miPic went down - total blackout. Deliberately sabotaged and the platform was offline for three months.

June 2021. We were preparing a Seedrs crowdfunding campaign, with £700,000 in investor pledges and 210 investors waiting. A £2 million opportunity for marketing that had taken 6 months to prep evaporated. Years that should have been spent developing the product became a fight to keep the company alive.

As miPic’s founder CEO I take my life’s project seriously, using my learnings as an architect to building miPic’s platform from scratch. We’d raised money via Crowdfunding including friends and family to venture capital 6-7 years working six days a week to make it happen.

Designing the product was easy, but I had never written a line of code. I never wanted to be a developer. When the platform was soft launched at Seed we depended on expert engineers to build and maintain until bugs and downtime began during the pandemic resulting in a 2021 blackout.

We spent years trying to recover without understanding the full extent of what had been done inside our own codebase.

Then Claude opened Pandora’s box

In February 2026, I connected Claude Code to our repositories and asked the AI agents to investigate. Within an hour, we were looking at a detailed forensic report.

The results were shocking.

We found a pattern across multiple deployments: core business functions disabled, error logging switched off, image processing broken and an automated print pipeline that could no longer do its job. Changes buried in the code had blocked revenue, fulfilment and new product launches.

From the outside, parts of the product still looked operational. Underneath, processes that should have been automatic depended on manual intervention. The product and marketing teams could no longer launch things themselves. One engineer knew how to keep those processes moving; the people hired afterwards were left working without the documentation they needed.

Three members of our team provided statements about unusual engineering behaviour prior to the downtime. Yet the consultant who had placed the developer had told me to “stop questioning the Devs deliverables.”

I’m responsible to shareholders, our customers and everyone who had backed me build the company. No stone wast to be left unturned, of course we are going to question the deliverables.

Four years of roadmap. Four weeks.

Once Claude could see the problems, I dived into a 3 week rabbit hole to rescue the platform code base. What started as seven AI agents, became 20 plus support from one ‘human’ developer and no new external engineering hires that previously cost £40-£100 per hour. I spent 360 hours across three weeks working with Claude day and night, reviewing the code, fixing needle in haystack breaks and building new product.

We repaired the system, shipped 13 features to production. The miPic PRO subscription went live within 24 hours. We released digital downloads, new print products and a new iOS app. Work that had sat on the roadmap for 2 years was finally reaching customers and users.

The refactor meant sabotaged page speeds came down from 20 to around 1.1 seconds. Infrastructure costs reduced by 60%. AI engineering cost less than $2,000 across February and March.

Four years of roadmap, delivered in four weeks would have traditionally cost $200k.

Getting my company moving again

I had always been deep in the details as an architect and close to the people building it. Claude gave me a way to carry that discipline further into implementation.

I could ask questions of the code, trace changes through our entire Git history and see what needed fixing and what was deliberately sabotaged. Decisions became working features while I was still close enough to question the result.

As a non technical founder who had spent years paying and trusting experts who attempted to silo knowledge of the system, agentic software development has changed everything.

Twelve years of work and four years of platform survival and planted bugs deserved more than another explanation for why we couldn’t ship.

We are building again.

I’ve shared the full forensic technical breakdown in detail here claude.mipic.co