Executive Summary
Steve Mays did not start smays.com as an AI project. He started it as a place to keep things: observations, links, jokes, book notes, arguments, worries, and passing enthusiasms. Over twenty-four years, that habit produced something newer technology could suddenly use: a dated, searchable, first-person archive large enough for machines to summarize, pattern-match, and reflect back to its author.
The interesting arc is not “blogger becomes immortal.” That would be too tidy, and the evidence does not support it. The better reading is more modest and more useful: Mays moved from blogging as personal memory, to blogging as public self-portrait, to blogging as a machine-readable body of evidence about attention over time.
Three shifts stand out:
- External memory: Early posts treat the blog as a way to preserve life as it is lived, with an immediate awareness that specificity creates both usefulness and exposure.
- Open-web selfhood: Through the 2000s and 2010s, Mays repeatedly contrasts the blog with committee prose and social feeds. The blog is where a person can speak in his own voice and leave durable evidence.
- AI-readable archive: By 2024–2026, he is asking LLMs to read, tag, summarize, and interpret the archive. The machine becomes less a publishing tool than an outside reader with unusual patience for volume.
The key observation is this: the most valuable thing here is not that Mays had unusually accurate predictions about AI. It is that he kept the raw material. Most people now generate fragments inside systems designed to forget, bury, or monetize them. Mays kept dates, tags, links, and ownership. That old-fashioned discipline aged well.
Essay
Mays began with a simple premise: he needed somewhere to put things.
That sounds smaller than it is. In 2002, he imagined a young person making regular blog entries for decades and ending up with “Your life online.” He connected the idea to his mother’s journals and immediately noticed the privacy problem: a life-record is useful because it is specific, and risky for the same reason. From the start, the blog was not only a publishing tool. It was memory with witnesses.
The long arc of smays.com is striking because it does not require much retrofitting. He kept going. The public archive still shows blogging-related posts in 2026, and artificial intelligence became a recurring category years before the current LLM boom. The neat irony is that he spent two decades building the corpus he would later ask machines to read.
At first, the blog is a place to write things down. Then it becomes a defense of human-scale publishing. In 2005, during the business-blogging moment, Mays endorsed the idea that “companies don’t blog; people blog.” The point was not content strategy. It was voice, conversation, and the failure of committee prose.
A month later, he argued that many people had always had something to say; they simply lacked an easy place to say it. He also noticed that bloggers often disclosed more of themselves online than they did at work or in ordinary life. That is an important shift. Blogging is not just broadcasting. It is selective self-exposure under the weak discipline of chronology.
Some of the claims Mays preserves are claims from other writers, not his own inventions. When he quotes Halley Suitt on blogs helping people remember thoughts, feelings, and values, the attribution matters. What is useful is not that Mays originated the idea, but that he found it worth preserving. The same is true when he quotes Seth Godin on blogging changing one’s posture toward humility, gratitude, and clarity; Mays adds practical testimony that five years of blogging had improved how he worked with clients and customers.
The philosophical hinge appears in late 2005, when Mays works through Dave Winer’s question about whether we are more than the stories we tell, and whether there is more to us than our blogs. That question keeps returning in updated costumes. If the blog is a story-machine, is it evidence of the self or merely performance? If it is incomplete, is it still more accurate than memory?
By 2019, Mays is closer to the latter view, or at least strongly sympathetic to it: if you care about your thoughts, keep them, because distant memory is not a trustworthy archivist.
Mortality enters early, with a dry little flourish. In 2003, he joked or half-planned an endowment to keep the site running after his death, including a video posted shortly after he was gone. This is not artificial intelligence yet. It is digital afterlife by scheduled post and stubborn hosting. But the instinct is there: the site should outlast the body, and perhaps speak after the author cannot.
The first explicit AI-readable version appears in 2008. In “Immortal Blog,” Mays imagined a future system reading thousands of posts, picking up his interests and writing style, posting after he dies, corresponding with friends, and reproducing audio or video. This is the key anticipatory moment. The blog is no longer just memory for Steve or evidence for readers. It becomes possible training material. The self is not uploaded, exactly. It is inferred.
That distinction matters. One way to read Mays’s later position is that it is not anti-AI. It is anti-amnesia, and increasingly wary of outsourcing the self. He wants machines to read the archive because no ordinary reader can easily hold the whole thing in view. In 2024 he wonders whether ChatGPT reading thousands of posts would improve interactions with him. The imagined AI is not only a ghostwriter for the dead Steve; it is an interpreter for the living one.
Three days later, he asks ChatGPT to review his tags, treating metadata as a rough portrait.
But he had prepared for that long before the tools were ready, at least in the practical sense of maintaining a searchable archive. In 2016, he described the blog as a flower bed: something that requires maintenance. He had deleted hundreds of posts and emphasized categories and tags because, once an archive reaches thousands of items, even the author may not know what to search for. The stated purpose was retrieval for himself. Later, that same retrieval architecture became useful to machine reading. The tags were not built for AI, but they were exactly the kind of structure AI could exploit.
His suspicion of social platforms fits this pattern. Twitter, Posterous, Facebook, and the rest appear as distractions or traps: useful in the moment, poor as memory. In 2010, he called the blog his first love. Later that year, he described the blog as the place where people could see what he thought and stood for, perhaps more clearly than people who had known him for years.
By 2017, responding to Dave Winer’s refusal to point to Facebook posts, Mays reaffirmed the open web and said he would have a blog until he died or could no longer maintain it, with an AI taking over if fortune allowed.
The contrast is simple enough to be almost impolite: social platforms remember what keeps the machine fed; the blog remembers what the person chose to keep. Mays makes the point again in 2026, when an AI-assisted widget helps surface old posts and he notes that social platforms care mostly about the immediate present.
The AI turn complicates the story without overturning it. In 2025, Mays compares the early AI era to Web 2.0 and says he is collecting traces as quickly as he can. The old archival impulse recurs under new pressure: something historically interesting is happening, so he stores his contact with it.
Later, in an AI-assisted analysis of posts about memory, he finds a gestalt, even a self-portrait he had not set out to paint. This is the mature form of the project: the archive looks back. Not mystically. Not with a soul. With retrieval, summarization, pattern recognition, and a great tolerance for repetition.
The attribution gets delicate here. In “When AI can ‘read’ my blog,” Mays asks whether AI might someday scan and analyze the full archive without manual upload. Claude frames possible uses: intellectual archaeology, buried strong work, unwritten books, contradiction mapping, and a queryable second brain. Those are Claude’s formulations, not Mays’s unaided prose. Mays’s editorial choice to publish the exchange tells us the possibilities interested him, but the voice should not be laundered into his.
He is also wary. In February 2026, he deletes his ChatGPT account over ads and engagement incentives while admitting LLMs are too useful to stop using. He notes that if such tools had existed in 2002, the blog would have been different. That sentence carries the real tension. AI can illuminate the archive, but it can also contaminate the conditions that made the archive worth reading. If the blog becomes too assisted or optimized, then the future “Steve” systems read may be partly a collaboration with the tools doing the reading.
Mays also published a Claude-written housekeeping post that formalizes this boundary. In March 2026, he created an AI-chats tag and let Claude explain the distinction between posts about artificial intelligence and posts containing conversations with AI systems. The useful taxonomy is adopted by Mays, but much of the language is Claude’s. Mays’s act was editorial: he accepted the distinction and published it.
That distinction is important for “The future of personal blogs,” where Mays explicitly says he is posting a ChatGPT take verbatim. He appears to like the frame, but the passage should not be treated as his solo authorship. The same caution applies more broadly: smays.com increasingly contains Steve, Steve quoting others, Steve prompting systems, and systems producing text Steve found worth preserving. The archive is richer for that, but less naive.
The culmination, at least for now, arrives in June 2026. Mays exports all posts, converts them to Markdown, and uploads them to Perplexity, Claude, and Gemini. He wants the full gestalt: what the posts say about who he is, who he was, and the journey between. He observes that, until then, no other entity had seen the whole thing.
That moment is revealing. The AI-readable self is not necessarily comforting. It can feel invasive even when the subject built the archive voluntarily, maintained it diligently, tagged it carefully, defended its platform, and finally handed it to the machines. External memory has become external interpretation. The result can feel like insight, judgment, flattery, distortion, or all four before breakfast. AI is efficient that way. Annoyingly efficient.
The lesson is not that Steve Mays achieved digital immortality. A blog is not a person. A model prompted on a blog is not a resurrection. smays.com is also not a complete record. It is selected, revised, tagged, pruned, and increasingly interleaved with other voices, including machine voices.
But that does not make it trivial. A persistent personal archive can become a durable substrate for self-interpretation, provided we remember that every archive is partial and every synthesis has an angle.
In practical terms, write 6,500 posts, wait twenty-five years, and eventually the machines will tell you what they think you meant.
More broadly, a person who keeps his thoughts in public, with dates and tags intact, gives both humans and machines a better chance of understanding the motion of a life. Not the whole life. Not the secret life. Not the final truth. But a record of attention over time. In an age of disappearing feeds, engagement traps, and synthetic fluency, that may be the difference between having an AI read you and having an AI replace the sound of you.