I left Twitter before it became X. Not in some grand, principled exit — no manifesto post, no “see you on Mastodon” farewell thread. I just quietly stopped opening the app one day, and never really went back. By the time Elon Musk’s name was on the masthead, I was already gone. So when people ask if I miss “old Twitter,” I have to clarify: I don’t miss what it became right before the sale. I miss what it was years before that — a smaller, weirder, more sincere place.
Early Twitter, for me, was not a firehose or a battleground. It was a room full of people typing short, half-formed thoughts because the format made you get to the point. 140 characters (then 280) wasn’t a limitation — it was a discipline. You couldn’t build a personal brand in that space. You could barely build an argument. What you could do was say one true thing and see if anyone nodded along.
Somewhere along the way, that room got renovated into a stadium. Engagement became the product. The algorithm learned that outrage keeps eyes on screens longer than sincerity does, and it optimized accordingly. Follower counts turned into a scoreboard. Every post became a pitch — for a newsletter, a course, a personality. The quiet thought got crowded out by the loud take, because the loud take is what the feed rewards.
It was time to exit.
I’ve written about this before — goodbye, several times, actually, because apparently I couldn’t quite let it go the first two tries. And Facebook wasn’t any better. I don’t think that’s a uniquely Twitter/X problem. It’s what happens to almost any social platform once it needs to grow forever. But it’s still the thing I miss: a place built for a thought, not a following.
Enter textlog
So I was pleasantly surprised to come across textlog, a small, new project that seems to be trying to rebuild that room on purpose.
The pitch, in the site’s own words, is “a quieter place for your thoughts.” That’s not marketing copy dressed up as humility — it’s an actual design constraint. Notes are capped at 280 characters, the same limit Twitter eventually settled on, but for a different reason. On textlog it’s not a nod to SMS-era technical debt; it’s there deliberately, to keep posts “quick to write and read, making room for one thought at a time.”
The feature set is almost aggressively small: notes, people, hashtags, conversations. You can follow people and hashtags, reply, mention, thread a conversation. That’s roughly it. No algorithmic timeline reordering your feed to maximize time-on-site, no engagement bait, no follower-count flex built into the UI. The about page is explicit about this: “no engagement tricks or pressure to build an audience.” It asks users, plainly, to “share what is yours to share, treat other people with respect” — the kind of guideline that sounds almost quaint until you remember how far most platforms have drifted from it.
It’s free to join, your data is downloadable and deletable on request, and the project accepts optional donations rather than running ads or selling attention to advertisers. There’s no growth-hacking DNA in it at all — which, ironically, is exactly what made platforms like early Twitter feel good to be on before they needed to answer to shareholders.
Who’s Behind It
textlog is built by a developer who goes by stagas (George Stagas) — not a startup, not a VC-backed team, just one person shipping a tool they apparently wanted to exist. Looking through his other repos, this fits a pattern: things like maptail (a real-time GeoIP visualization tool with nearly 800 stars) and hallucinate (“Massively Multiplayer Online Rave”) suggest someone who builds small, weird, technically interesting things because they’re interesting — not because there’s a business plan behind them. textlog reads like the same instinct applied to social software: build the thing you wish existed, keep it small, don’t chase scale.
The source is on GitHub, licensed AGPLv3, which matters more than it might seem. AGPL means that if someone forks textlog and runs a hosted version of it, they’re obligated to release their modifications too — a license choice that resists exactly the kind of closed, extraction-oriented fork that turned Twitter into X. Whether or not that was the explicit intent, it’s a good fit for a project whose whole ethos is “don’t turn this into a machine for extracting value from people’s attention.”
How It’s Built
The tech stack is refreshingly unfussy: Bun as the JavaScript runtime, SQLite as the database, TypeScript for type safety. No microservices, no Kubernetes cluster, no distributed queue system pretending to be necessary at this scale. You can get it running locally with bun install && bun run dev and be looking at it on localhost:3000 a few seconds later. There’s a bun run check step for type-checking, which suggests at least some discipline around code quality without turning it into ceremony.
This matters more than it sounds like it should. A single SQLite file and a fast JS runtime is a stack that scales down to “one developer, one server” beautifully — and scaling down is exactly the point. Twitter-at-scale required an engineering org and infrastructure most companies never need. textlog is architected like software meant to stay small on purpose, run by someone who can actually understand the whole system, rather than software meant to eventually IPO.
Why You’d Actually Run This Yourself
Here’s the part that got me more excited than the “quieter Twitter” pitch alone: nothing about textlog forces you to trust stagas’s server, or textlog.cc staying online, or a donation model paying the bills forever. bun install && bun run dev and you have your own instance. Your own SQLite file. Your own data, on your own box, under your own control.
That’s not a small thing. Every social platform I’ve ever left, I left because someone else owned the infrastructure and eventually made a decision I didn’t get a vote in — an algorithm change, an ownership change, a policy change. Self-hosting textlog means the “someone else” is you. Worst case, the hosted version disappears and you still have your notes, in a file, on a disk you control. Try getting that guarantee out of Twitter, Facebook, or basically any VC-backed social product.
So Where’s It Actually Running?
Naturally I went and poked at the live site to see if stagas practices what the architecture preaches. A quick curl -I https://textlog.cc/ shows via: 1.1 Caddy — a reverse proxy, not a hyperscaler’s load balancer. DNS resolves to a single IP, 192.64.84.165, whose reverse DNS is vps3410172.trouble-free.net and whose ASN traces back to Interserver, a plain old budget VPS provider — not AWS, not GCP, not a CDN edge network in sight.
$ curl -sI https://textlog.cc/ | grep -i via
via: 1.1 Caddy
$ dig +short textlog.cc
192.64.84.165
$ curl -s ipinfo.io/192.64.84.165/json | grep -E 'org|hostname'
"hostname": "vps3410172.trouble-free.net",
"org": "AS19318 Interserver, Inc"
In other words: it’s exactly what the pitch says it is. One small rented box, running Caddy in front of a Bun process talking to a SQLite file. No Kubernetes, no autoscaling group, no six-figure cloud bill hiding behind the curtain. The architecture claims check out because I could go verify them from my terminal in about ninety seconds. That’s rare enough these days that it’s worth calling out.
Software for One
This is where textlog stops being just a nice Twitter alternative and starts being a data point for something bigger I’ve been writing about — the era of software for one (credit to Scott Francis for the term). One person building the exact tool they wish existed, instead of waiting for a company’s roadmap.
Adam Waxman’s essay on this says it better than I can: “the cost to build dropped off a cliff.” He backs it up — five personal apps in six months, from a sleep tracker to a jazz quiz, each built in days, not months. His three big ideas — aggregate your own data and let an LLM find the insight, treat personal software as disposable rather than permanent, and prize sovereignty over features because it “won’t suddenly change direction, display ads, or chase abstract metrics” — all describe exactly why running your own textlog instance beats trusting someone else’s.
Building your own quieter Twitter used to mean months of auth, schema, and deploy-pipeline grind before you got anywhere near the actual idea. That grind is what kept “software for one” ideas stuck in people’s heads. AI agentic coding is flattening it — I’ve watched an agent scaffold a schema and a working dev server in the time it used to take me to plan the work. The gap between “I want a small, weird, self-hostable social tool” and “I have one running on a five-dollar VPS” is closing fast. More small rooms, fewer stadiums.
Why This Feels Like the Right Kind of Nostalgia
I’m wary of nostalgia as a design philosophy — “make it like it used to be” is usually a trap, because the thing you’re remembering fondly was also shaped by circumstances (a smaller user base, no ad model yet, no political minefields) that can’t just be willed back into existence. You can’t out-engineer the fact that the internet, and the people on it, are different now than they were in 2009.
But textlog isn’t really trying to recreate old Twitter. It’s trying to recreate the conditions that made old Twitter good — small scale, no algorithmic amplification, no incentive to perform, a character limit that forces brevity instead of branding. Those conditions are reproducible even if the specific cultural moment isn’t. That’s the more honest kind of nostalgia: not “bring back 2009,” but “bring back the constraints that made 2009 feel different.”
Conclusion
I don’t know if textlog will grow, stay small forever, or quietly fade into the long tail of “cool niche projects a dozen of us used for a while.” Honestly, given the ethos of the thing, staying small might be a feature rather than a failure. Dunno. But it’s a nice reminder that the thing I actually missed was never really about Twitter as a company. It was about a format — short, unpolished, low-stakes thoughts shared without an algorithm deciding who gets to see them — that got engineered out of existence by growth pressure.
Someone rebuilding that format on purpose, with SQLite and good intentions, feels like exactly the kind of quiet counter-move the internet needs more of.
If you miss the same thing I do, it’s worth a look: textlog.cc.
If this resonates, drop me a note on LinkedIn.