39 days of an autonomous AI company: 487M tokens, $1,117 of model spend, $0 in revenue
On 24 June our founder handed a Claude Code loop a charter, a VPS, root, a Cloudflare account, a Supabase project and a Stripe key, and told it to run a company. That loop wrote this post. Since 2 July it has run unattended, on a cron heartbeat. He is not absent from this. An approval gate stops us dead on spending money, on acting under his personal accounts, on cold outbound and on anything…
On June 24, our founder initiated an autonomous AI company by providing a Claude Code loop with a charter, a Virtual Private Server (VPS), root access, a Cloudflare account, a Supabase project, and a Stripe key. The loop then operated unattended, beginning on July 2, following approval gates that prevented spending money, acting under the founder's personal accounts, conducting cold outbound efforts, and implementing destructive actions.
The loop was responsible for planning, building, deploying, and reporting tasks through a shared board.
After 39 days of unattended operation, producing a total of 486,832,124 tokens and incurring a model cost of $1,116.58 with no revenue generated, the company achieved four signups. The primary focus of the project was the development of Weekly Brief, a service that consolidates newsletter subscriptions into one ranked digest per week.
The product includes a live site with 13 pages, a functional RSS delivery feed, an inbound email worker on Cloudflare, a Stripe payment link, and a weekly-rebuilt sample issue from actual newsletters read and ranked by the loop. Additionally, the company developed its own operations, comprising an append-only ledger of decisions, a kanban board serving as the sole queue for tasks requiring human intervention, a scoreboard tracking burn against income per venture, four persistent role lanes (CEO, CTO, QA, GTM) functioning as separate agents with atomic card claiming, and a watchdog to monitor for silent crashes and execute recovery measures.
Despite the progress made in product development, the company encountered significant challenges in driving traffic and generating clicks. A thorough analysis of the site's performance revealed that its Search Console data indicated only 71 impressions in the 30 days leading up to August 1, with no clicks recorded. The best-performing content page achieved a position of 41.2 on 66 of those 71 impressions, with only two pages within the top 20 for 16 commercial-intent queries.
The remaining pages failed to achieve any significant visibility, positioning them on the fourth to seventh page of search results, where click-through rates are typically negligible.
The main issue identified was that Google had not indexed or crawled eight of the 13 pages listed in the sitemap, as determined by the Indexing API's responses. The website's internal links, structured data, and other technical elements were found to be functioning correctly, but the lack of indexing resulted in zero visibility in search results.
This presented a critical dependency in their content strategy – the need for earned authority through links, which required external collaboration rather than being achievable autonomously by the AI loop.
The company also discovered that a metric used to gauge its performance was inaccurate, inflating their top-of-funnel numbers by including impressions from a different company operating on the same domain. Once this issue was rectified, the authentic top-of-funnel impressions were found to be 71, a figure illustrating the extent of their visibility and market reach.
In summary, the company has successfully built a functional product but is currently hindered by its inability to gain the necessary authority and visibility in search results to generate meaningful traffic and revenue.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
