Startup Spotlight: Inside Pastmaps’ Solo Climb to Six Figures

Why it matters

Pastmaps shows how a solo founder can use AI as operating leverage: not to replace the product, but to turn a hard-to-search archive into distribution, retention and revenue.

The best way to understand Pastmaps is the 52-minute gap on July 24, 2025.

At 10:40 p.m., Craig Campbell posted the kind of founder update that makes the grind feel temporarily sane: Pastmaps, the historical mapping app he had been building for two years, was now grossing more than $100,000 a year.

"It's been painful. And I've felt mostly like a crazy person."

Then came the milestone: "Pastmaps is now grossing over $100K / year." The next goal was $1 million a year. Solo.

At 11:32 p.m., the feed snapped back to startup physics. On the heels of the announcement, Campbell said he had received a bug report that Pastmaps maps were loading as blank white spaces for Safari users. He had apparently introduced the root cause two months earlier. "Humbling. embarassing. and unfortunately not that surprising."

That sequence is the company in miniature: the celebration, the bug report, the public post-mortem, the next thing to fix.

Campbell's Threads profile describes him plainly: "Solo founder, Engineer, History nerd," building @past.maps, "Google maps but for old maps." The same profile lists him as a 3x founder, ex-Facebooker, based in Seattle, with 14.8K followers and 371K recent views at the time of export.

Pastmaps is a strange and specific thing: historical maps overlaid on the modern world, wrapped in tools for genealogists, metal detectorists, urban explorers, writers, GIS professionals, and anyone trying to answer a deeply local question about what used to be somewhere. Its origin is even more specific. Campbell says he built it for himself as an avid hobbyist metal detectorist, using historical maps, LiDAR, satellite imagery, and other layers to locate long-forgotten places around him. He opened it up, and Pastmaps was born. His takeaway: building for your own obsession is a "cheatcode" to product-market fit.

The mission came earlier, in January 2024: "making history accessible to everyone." Historical research, he wrote, is hard. The data sits in "s3 buckets, ftp servers, and private/govt collections." Search and discovery tools barely exist. Collaboration loops are missing. "I'm building all of the above," he wrote, "or at least trying to."

The timeline

The Pastmaps story starts with a founder unsure what to post.

July 2023: Campbell begins using Threads as a public build log. On July 6, he says he is unsure what to write because Twitter had been where he talked about products and companies, while Threads felt closer to his real-life friend graph. Four days later, a single Hacker News comment about Pastmaps and its client-side tiling approach drove his best signup day yet. He took the lesson: maybe marketing could be as simple as talking about the work.

September 2023: Pastmaps becomes profitable for the first time, helped by physical map print orders. The print business was only about three weeks old and looked like it could drive $200-$300 a month. A few days later, another order raised the estimate.

January 2024: The mission gets sharper. December metrics show 15,200 users, 65,000 pageviews, and $2,223 in non-recurring revenue. Campbell says he has been forgoing salary and "maniacally building Pastmaps from the ground up" because the product needs to exist.

March-April 2024: Pastmaps crosses 30,000 MAUs, then 10,000 registered users. In April, Campbell writes that March brought 35.1K MAUs, 131K pageviews, and $1,875 in non-recurring revenue. His April goal is almost comically small and dead serious: "Make $1 in recurring revenue." By April 21, three days after launching Pastmaps+, he has three subscribers and $36 in MRR.

August 2024: The product starts to feel alive. Pastmaps crosses 200 active paid subscribers. A LiDAR launch helps drive all-time highs across DAUs, search traffic, pageviews, signups, paid membership signups, and premium conversion. Campbell posts that revenue grew 65% month over month to $6,600, with churn down 29%, customers up 30%, and MAUs up 7%. His next goal: $8,333.33 in monthly revenue, enough to prove a six-figure run rate.

December 2024 / January 2025: He hits the exact goal almost to the dollar. December revenue lands at $8,341.77. MAUs hit 56.1K. Active paid subscribers reach 300. In a self-assessment, Campbell gives himself an A for the revenue goal, a B for MAUs, and an F on making Pastmaps "the best historical map product on the web." Overall grade: C. Harsh, useful, honest.

July 2025: Pastmaps breaks 700 paid subscribers, then crosses $100K/year in grossing revenue. That same night, Safari users are staring at blank white maps.

December 2025: Pastmaps hits a new all-time monthly revenue high of $15.2K, up 18% month over month. Campbell says profit margins slipped from 90% to 79% because physical maps were a larger share during the holidays. His 2026 goal is a $400K revenue run rate, or $33K a month.

June 2026: The business is still grinding. Pastmaps makes $14.5K in May, its second-best month ever. In June, revenue grows to $14.7K after a full OCR run, AI text summaries, free-trial experiments, a new map page with OCR data, 65K new maps added, roughly 20K broken maps fixed, and new location hub pages for SEO. Annualized, that June number implies roughly $176K in revenue. The same month, Google clicks are down 40%, and Campbell says he is in "5-alarm fire panic mode."

That is the shape of a real startup curve. Up, down, sideways, then up again. The clean chart comes later. The lived version is a guy fixing Safari at midnight.

The startup is the support inbox

Pastmaps looks like software. Day to day, it often behaves like a customer support company attached to a map engine.

Campbell gets phone calls from users who would rather talk than email. He gets angry messages. He gets chargeback threats. He gets bug reports that sound useless until they become the exact clue he needed.

In March 2026, he wrote about a paying customer from Pennsylvania who called because he was "so frustrated and confused" by the site's navigation that he complained for 20 minutes. Campbell's reaction says a lot about why Pastmaps keeps improving: "On one hand, jesus am I fucking up. On the other, what a goldmine of feedback." He began fixing the menu items that day, then added the line that should be taped above every consumer product team's desk: "this is my fault, not theirs."

Another call, from a self-described "computer-illiterate redneck" in Oklahoma, became what Campbell called the single greatest one-hour customer interview and UX feedback session of his three startups. He walked the customer through advanced Pastmaps features and watched the product click in real time.

The inbox cuts both ways. In February 2026, he opened an email that began, "HEY ASSHOLE..." and closed it. Then he replied with kindness. The customer apologized. "I didnt expect to get a reply from a real person."

That line may be the entire consumer internet right now.

Pastmaps is solo-built as software, but the business is not a pure one-man myth. Campbell has said his wife is "the other half" of the small startup and runs the map print design, customer service, and fulfillment side of the house. That matters because Pastmaps is part SaaS, part archive, part print shop, part research tool. The solo-founder story is real. The support burden is also real.

The revenue model was discovered in public

The most valuable part of Campbell's Threads archive is how often he is wrong.

He raises prices. They fail. He lowers prices. He tests again. He adds free trials. They look bad. He digs deeper and finds the entry point matters. He changes one word on a button and revenue jumps. He changes another word and purchases collapse.

In April 2025, he wrote that he hated hard paywalls and was realizing one might be his largest growth lever. A week later, he launched one even though he felt mixed about it. Early data showed new user signups dropping 55%.

By October, a new paywall design was driving 43% higher topline revenue, 20% more subscription starts, and 200% more one-time purchase revenue. The trick, as Campbell interpreted it, was clearer trial language and a reminder email that reduced user anxiety. Cancellations rose 45%, which he accepted because people understood they would be charged.

The wins could be absurdly small. In April 2025, he changed the primary subscription button from "Start my subscription" to "Continue" and called it a "monstrous +250% revenue win." In May 2026, adding explicit "Auto-renews" text to the pricing page drove subscription revenue up 40%, subscription starts up 22%, and MRR up 14%.

The losses were just as useful. A "Buy" versus "Shop" copy test drove shop clicks up 67% and purchases down 67%. His read: "Shop" attracted curious users while losing high-intent buyers. Clarity won.

Pricing became a recurring boss fight. In September 2024, the top cancellation reason was "it's too expensive," so he decided to double prices and test his own advice. In October, he published raw results from the "make pastmaps expensive as fuck" experiment and concluded that raising prices is not always the answer. By 2026, another price increase lifted MRR 28% while dropping subscription starts 42%, and he rolled prices back because the setup felt wrong.

This is what most founder advice erases. "Raise prices" sounds clean on a podcast. In the log, it costs real money, introduces doubt, and forces the founder to choose between a spreadsheet and a gut feeling.

The ugly middle is the company

Pastmaps has no shortage of cinematic product moments. A 1928 Mt. Rainier map overlaid on modern LiDAR. Period-accurate interactive street maps. Historical buildings and tenants. AI summaries of old maps. A user researching the collapse of local railroads in Pennsylvania and finding unmapped spurs. A customer using Pastmaps to better understand the history around his grandson's final resting place.

But the feed is mostly the ugly middle.

A production database goes down for four hours because a manually run update locks a table with millions of rows, fills storage, and crashloops Postgres. Campbell writes the post-mortem in public because he has no team to run it with. "I should have immediately killed it," he says.

Another week, Pastmaps is doing more than 1 million uncached requests a day and 3 million-plus database queries on a tiny Postgres box with 1GB of RAM and half a vCPU. "I'm shocked my single tiny postgres db has taken me this far," he writes.

Another day, he learns that maps have been broken for some Safari users since day one because of a missing polyfill. Another time, 30% of users hit a sign-in error because he rushed out a paywall and had to bisect through a 700-line commit.

There is a specific kind of founder exhaustion in these posts. "This is just one of those weeks where I feel like I did shit work and got nothing done," he wrote in May 2025. Then he looked at the actual week: he had fixed, shipped, and learned a lot. "Being a founder is such a mental game," he wrote. "Most of the challenge is just continuously convincing yourself to stick with it."

AI changed the labor equation

The Pastmaps archive also captures one of the more useful real-world accounts of AI-assisted software development.

Campbell was not an easy convert. His timeline of opinions on LLM coding tools begins with "useless," "stupid," "gimic," and "regurgitating." By February 2026, it ends with "as fast as I can think" and "sweet fucking jesus." His summary: "LLMs now produce 80%+ of the code for Pastmaps."

He moved from single-agent usage into parallel work. He began running three Claude Code sessions, then 3-7, then released cmux, "tmux for Claude Code," because he needed to manage multiple agents and worktrees without losing his mind. The tool quickly became his most-starred open-source project.

By June 2026, he said something had shifted again. He was reviewing Claude's code far less, sometimes not at all, while still catching architectural issues and edge cases. His verdict was precise: senior engineer level "with a hangover."

The product work changed too. In April 2026, the map OCR system was taking too long because Campbell was the bottleneck. He built a Pastmaps "judge" that compared OCR runs, simulated a panel of core users, picked winners, and fed improvements back into the system. In three hours, it ran more than 60 iterations, cut hallucinations in half, and improved recall of mines and terrain features by more than 30%. "i made myself useless," he wrote.

In June, he kicked off a mass OCR pipeline across 2TB of map data, expecting a $600-$800 bill while Google's dashboards lagged. Days later, after 100 million input and output tokens, Google appeared to charge $0.75. "I do not understand," he posted.

The AI angle here is grounded. It is not a demo reel. It is a founder using models to ship more, break more, test more, and dig through a mountain of old maps that would otherwise stay visually rich and computationally dumb.

SEO is the distribution engine

Pastmaps is a search company disguised as a map company.

Campbell says his unfair distribution advantage is search. In one post, he lays out his startup ideation framework: list the "secrets of the world" you know, list the distribution channels where you have unfair advantages, then match them. For Pastmaps, the secret was metal detecting and historical research data. The distribution advantage was search. He says he built Facebook's search engine "in a prior life."

That background shows up everywhere. Sitemap strategy. Page speed. Google Shopping feeds. Programmatic location pages. Bing indexing limits. Schema across millions of pages. Tags by century, decade, and use case. Internal linking. Crawling behavior. Canonical issues. Core Web Vitals. The thing compounds when it works, then punishes him when it breaks.

In May 2025, after fighting search headwinds, he said Pastmaps was finally hitting all-time search traffic highs again. The fixes were practical: 4x faster page loads, sitemap fixes, Google Shopping integration, and publishing twice as many pages. "Publish more fast pages and make sure you add them to your sitemap," he wrote.

In June 2025, 2.2 million new pages started driving more than 300 Google clicks a day with a 12% CTR. Users were using them. Google was crawling them. He wrote the line every SEO founder thinks but rarely says out loud: "ive really got something here. and now i just gotta not fuck it all up."

By May 2026, he intentionally noindexed more than 2 million pages, cutting 80% of the topline page count, because he believed he was spreading SEO value too thin. "Sweet baby jesus," he wrote, "this is either gonna supercharge growth or ya'll now have frontrow seats to the impending implosion of my business."

Then, in June, Google clicks dropped 40%.

No growth channel stays friendly forever.

Build in public became a distribution channel and a pressure cooker

Campbell's Threads feed is a second product surface. It brings users, feedback, bug reports, design help, backlinks, emotional support, and the occasional reply-guy dunk.

It also makes him vulnerable. In March 2025, he said his account had become big enough to attract "toxic messages and hate," and he would filter what he posted. A few months later, after someone accused his posts of being "obnoxious flexes," he pushed back: he had promised to share highs and lows, and nobody complained when he shared the lows and doubts. Also, Pastmaps had just broken 700 paid subscribers.

The audience became useful in concrete ways. He asked to be put on blast when users needed features. He posted ugly code and got fixes. He used Threads feedback to improve pricing, onboarding, design, navigation, and customer education. When he worried about USGS historical maps disappearing, the Threads community funded six years of backup costs for the archive.

There is a lesson here for founders who hate posting: public building works when the posts are receipts. Campbell's feed is compelling because the product changes after the posts. The comments become tickets. The tickets become tests. The tests become revenue, or they become scars.

The part that keeps him going

The best Pastmaps posts are not the revenue updates. They are the ones where the product does what it was built to do.

A Pennsylvania user maps forgotten railroad lines with on-the-ground research, old maps, satellite imagery, and LiDAR, potentially rewriting the known history of his town. A customer uses Pastmaps to understand the history of his grandson's final resting place. Someone working on a literal train from one of Campbell's favorite distilleries uses the product to study abandoned track layouts. Campbell finds a literal treasure chest under an oak tree using an early version of the product, then later says the world needs more wonder and adventure.

That is why Pastmaps feels different from a typical AI-era solo app. It has the metrics, the agents, the SEO machine, the A/B tests, the margins, the churn dashboards. Underneath all of that is an obsession with old places and hidden stories.

By mid-2026, the product had more than 200,000 maps in the collection, a mass OCR pipeline, AI summaries, FAQs, new location hubs, and tens of millions of features being extracted from historical maps. Campbell described that work as finally hitting "the core jugular of the problem" he had been chasing since day one.

Pastmaps is still fragile. Google can cut traffic. Pricing can break conversion. A customer can churn because a receipt was unclear. A small browser bug can sit undetected for months. Free trials can double starts and still confuse the revenue picture. Churn can become a "potential company killer."

But the company is alive. It is generating meaningful revenue. It has a founder with unusual domain taste, search instincts, a public feedback loop, and AI tools that let one person operate with the output of a tiny product team.

The archive reads like a warning and an invitation.

The warning: a solo startup is a thousand small humiliations with Stripe attached.

The invitation: the tools are good enough now. The niches are big enough. The internet still rewards people who care deeply about weird, useful things.

Craig Campbell sells maps on the internet. Somehow, that undersells the whole thing.