Alexis Bouchez

~ / startup analyses / Newsletters You Could Actually Run For Forty Years

Newsletters You Could Actually Run For Forty Years

On March 21 you wrote in your journal: "I need to find which is the piece of software I should obsess over for the next 30 years." You asked the same question on February 19 about a codebase. You keep circling it. Forty years is just the honest version of that question.

A newsletter can be the same bet. The trick is not picking a topic that sounds "sustainable." The trick is picking a topic where issue number 2,080 still feels like the easiest thing in the world to write. Neil Sloane started the OEIS in 1964 on punched cards and he is still adding sequences in 2026. Etymonline is twenty five years old and Douglas Harper still updates it. Those are the reference points, not the one year newsletter that gets acquired by Morning Brew.

So: no thinkpieces. No "weekly roundup of what is hot in X." Those die. Every single one of them dies. Instead, every idea here is an obsessive catalog: one new entry per issue, adding to a permanent archive. The newsletter is the distribution layer; the archive is the asset. After ten years you have an encyclopedia nobody else has. After forty you have a small institution.

1. The filter I used

Before listing anything I threw away ideas that failed any of these:

  1. Does it match an obsession you already have on the record? I checked your research directory. If there is no corresponding rabbit hole already in there, I did not suggest it.
  2. Will you still want to write it on a bad week in year twelve? This kills almost every "trending topic" newsletter. Hot takes rot. Catalogs compound.
  3. Can one person run it with no team, no VC, no burnout sprint? You said on March 21 you hate working for someone else. Any format that requires an editor, a community manager, or a scraping infrastructure team is out.
  4. Does each issue add to a permanent archive, not just an inbox? If the only place the issue lives is in a Gmail "promotions" folder, it is not a 40 year asset.
  5. Does it avoid the things that have already exhausted you? No SaaS, no cold outreach at scale, no "here is a tool you can buy" affiliate angle. You tried those. You said on April 7: "I hate working for someone else too much." These suggestions respect that.

2. 1. Oubliés: un penseur oublié par semaine

Language: French. Cadence: weekly. One forgotten thinker per issue.

You have at least eight research files on underground and forgotten thinkers. forgotten-thinkers-iran, underground-thinkers-living, neoreaction-unknown-thinkers, byzantine-thinkers, presocratic-philosophy, ussr-soviet-thinkers, underground-thinkers-france-america-england-ireland, nick-land. That is not a topic. That is a canon you are already writing from memory.

Format: every Sunday, one profile. 800 to 1,200 words. Name, dates, the one idea that matters, why they got dropped from the mainstream story, where to read them today, a single quoted passage. No thinkpiece framing. No "what this means for our times." Just the thinker and the receipt.

The arithmetic: 52 profiles a year, for 40 years, is 2,080 thinkers. There is no French language reference that does this. The nearest things in French are Wikipedia (too formal, no narrative) and a few academic blogs (dead by 2014). You would become the default search result for hundreds of names over a decade.

Why you will not quit in year three: you already cannot stop reading about these people. On March 21 Claude noted that your most alive moments are when you are "building technical things for yourself." The prose equivalent is writing about people who fascinate you. You are not performing an interest; you are taxing an existing obsession.

Monetization honesty: Patreon and paid tiers, not ads. Target 2,000 paying readers at 5€ per month by year five. That is 10,000€ per month, solo, with zero sponsors to chase. Closer to what Tim Carmody does with Amazon Chronicles than what TLDR does.

What would kill it: trying to be comprehensive in year one. Pace is the whole game. One per week. Forever. If you miss a week, skip, do not apologize.

3. 2. The Artisan Developer

Language: English. Cadence: weekly. One solo developer who built one thing for ten plus years.

This is the newsletter version of your "30 year software" question. You already know the archetype: Fabrice Bellard (QEMU, FFmpeg, tcc), Neil Sloane (OEIS), D. Richard Hipp (SQLite), Daniel Stenberg (curl), Linus (Linux), Jamis Buck (Ruby on Rails, then mazes). Each of them obsessed over one codebase for decades. You want to be one of them. Writing about them every week is how you build the mental model for what it actually looks like from the inside.

Format: one profile per week. The person. The tool. The pivot points. The years when they almost quit. What they ate during those years. Who paid them. What the tool looks like at line 47,000 of the commit log. Real interviews where possible; public record where not. Think Stripe Press energy, but one person, monthly frequency, obsessed.

The arithmetic: 52 per year. By year ten you have profiled every living developer who built something alone for a decade. By year twenty the profiles double back on the same people at different stages of their life. By year forty the newsletter itself becomes a historical record of how software got made before AI ate most of it.

Why it fits you specifically: you cited Linus choosing Linux in your journal. You keep saying "obsess over one piece of software." Writing a weekly profile of people who did exactly that is the cheapest way to internalize their shape. You will also accidentally become the person every young indie hacker emails when they want to know "how do I become Fabrice Bellard." You will own that position.

Monetization honesty: this is the most fundable of the seven. Every devtools company on earth would sponsor this. Sentry, PostHog, Fly, Turso, Railway. One sponsor per issue at 1,500€, 50 issues a year, is 75,000€. Keep the sponsor in a single clearly marked block at the bottom. Do not let sponsors pick the profiles. Ever.

What would kill it: letting it become a trends newsletter. The moment you add a "this week in devtools" section, it is dead. Stay single subject.

4. 3. Katabase: One Descent Per Week

Language: English or French, pick one. Cadence: weekly. One descent into the underworld per issue.

You have a file literally called katabasics.md. The descent myth is one of the oldest structures in human literature and you are already pulling on the thread. Orpheus. Inanna. Odysseus in Book 11. Dante. Virgil fetching Aeneas's father. The shaman's journey. Carl Jung. Modern versions in literature and cinema. Every culture has them. Every era rewrites them. You will never run out.

Format: one descent per issue. The text or film. The descent. What the protagonist is looking for. What they bring back. The one thing this version does that no other version does. 1,000 to 1,500 words. Quoted passages when possible. No "seven life lessons from Orpheus." This is not LinkedIn.

The arithmetic: katabases are everywhere once you look. Ancient epic, Gnostic cosmology, Sufi poetry, Norse myth, shamanic practice, Kabbalah, Dante, Milton, Blake, Melville, Rilke, Cormac McCarthy, Tarkovsky, Lynch. Forty years of one per week and you are still finding new ones.

Why it fits you specifically: your research directory reads like someone who keeps circling myths and mystical texts. plotinus-enneads-analysis, ancient-mysteries-analysis, presocratic-philosophy-analysis, sol-invictus-analysis, iliad, and then katabasics itself. You are building the same reader list whether you want to or not.

Monetization honesty: this is the hardest to monetize cleanly. No B2B sponsor fits. Patreon tiers, paid archive, and eventually a book. A book deal after ten years of this is almost guaranteed because no such book exists.

What would kill it: trying to be academic. This has to be readable. Short sentences. A descent per week as a working writer would read it, not as a PhD would write it.

5. 4. Infra Hebdo (French DevOps Weekly)

Language: French. Cadence: weekly. One infrastructure tool, one pattern, one hire, one incident post mortem, one obscure config trick.

You already wrote the playbook for this in french-niche-newsletters-playbook.md. Revenue model is solved: 500€ per sponsored slot, two slots per week, 4,000€ per month solo within eighteen months. The French professional developer market is large enough (hundreds of thousands of people) and underserved enough (almost zero French language curation) that the math is not speculative.

Format: five short sections per issue, always the same structure. Un outil, un pattern, un poste qui recrute, un post mortem, une astuce. Never break the format. Never do a "special edition." The boringness of the format is the feature: readers can consume it in three minutes on the train and know exactly what they are getting.

The arithmetic: 52 issues a year times five items is 260 things per year. After ten years that is 2,600 curated items, indexed and searchable on your own site. This is where you accidentally build a niche encyclopedia without calling it one.

Why it fits you specifically: your blog posts on Docker Go SDK, PostgreSQL with pgx, HTTP error handling in Go are exactly the register this newsletter needs. You already write this voice. You just have not been distributing it in newsletter form.

Monetization honesty: the cleanest path of the seven. You have 10,000 French developer subscribers within three years if you launch with a single good issue and show it to your network. After that, two sponsors per week, no drama.

What would kill it: getting bored and "experimenting" with format. The playbook says do not experiment. The playbook is right.

One caveat: Infra Hebdo is the only newsletter on this list whose topic might feel less personal in year twenty than in year one. I would still recommend you launch it, because it is the one that pays for the other six. You can always hand it off to a co editor in year fifteen.

6. 5. L'Hérésiarque: One Heresy Per Week

Language: French. Cadence: weekly. One religious heresy, schism, or forgotten sect per issue.

Your research directory is a map of this obsession already. islamic-criticisms-of-christianity, jewish-criticisms-of-christianity, pagan-criticisms-of-christianity, orthodoxy-catholicism-comparative-theology, protestant-traditions-analysis, judaism-branches-analysis, islamic-branches-analysis, hinduism-traditions-analysis, bible, bible-sexe, evangiles-sexe, plotinus-enneads-analysis. You have been quietly building a comparative theology lab in your research folder for months.

Format: one heresy, sect, or schism per issue. When. Who. What they actually believed (not what the winning side said they believed). Why it was suppressed. Whether a trace of it still exists today. Close reading of one primary source if possible. 1,200 words max.

The arithmetic: the set of suppressed or forgotten religious movements across 3,000 years and every major tradition is effectively unbounded. Gnostics, Cathars, Manichaeans, Bogomils, Paulicians, Waldensians, Mutazilites, Ismailis, Druze, Alawites, Yazidis, Karaites, Sabbatians, Doukhobors, Jansenists, Molinists, Old Believers, Ranters, Diggers, Muggletonians, and we have not touched Hindu, Buddhist, or Chinese traditions yet. Forty years is not enough to run out.

Why it fits you specifically: you are drawn to marginalized canons. This is the religious version of Oubliés. The two newsletters rhyme without overlapping; Oubliés is about individuals, L'Hérésiarque is about communities and doctrines.

Monetization honesty: paid tier only. No sponsors fit. But this is the one that an academic press eventually turns into a book series. Ten years of L'Hérésiarque plus a decent university contact equals a multivolume reference work with your name on it.

What would kill it: taking sides. Write the heresy from the inside, the way the heretics themselves would recognize. If you start apologizing for orthodoxy or for heresy, both kinds of reader leave.

7. 6. Codex: One Manuscript Per Week

Language: French. Cadence: weekly. One primary source text per issue, with one passage translated or close read.

A slower, even more specific cousin of Oubliés and L'Hérésiarque. Instead of a person or a movement, each issue is one text. A Byzantine chronicle. A Persian treatise. A Coptic gospel. A medieval recipe book. A Tibetan terma. A Mutazilite letter. You find it, you describe the manuscript, you translate or quote one page, you explain why the text matters, you link to where to read the rest. Done.

The arithmetic: the world's unread primary sources number in the hundreds of thousands. A weekly newsletter can run for centuries on this material. The main constraint is your own reading speed, not the availability of texts.

Why it fits you specifically: you wrote three "future of academic fields" research files in a row (future-of-history, future-of-anthropology, future-of-religious-studies, future-of-literature). You have an instinct that academia is abandoning ground you find important. This newsletter is the nondestructive way to take that ground for yourself without enrolling in a PhD.

Monetization honesty: this is the slowest to monetize and the most durable. No sponsors. A small paid tier. A Substack that becomes the reference that grad students quietly cite. Your reward is not revenue; it is that you become, over twenty years, a person who actually knows things. The revenue from the other newsletters pays for this one.

What would kill it: scope creep. One text per issue. Not two. Not "a thread of related texts." One.

8. 7. Le Carnet de l'Artisan

Language: French. Cadence: weekly. One week of your own craft, documented.

This one is different from the other six. It is not a catalog of other people. It is a catalog of your own work. One issue per week, always the same format: what you built, what you read, what you threw away, one paragraph on a problem you are stuck on, one paragraph on a small thing that worked. It is your public journal, edited for other people, published on a schedule.

The model is already in your head. You said on March 4 (per Claude's journal note) that you were writing your entries more for AI than for humans. Le Carnet is the human facing version of the same practice. It has a natural cap of one page per issue. It has no research cost. It has no sponsor conflict. It compounds in the weirdest way: after ten years of weekly issues, the archive itself is the product. Readers binge it like a novel.

The arithmetic: 52 issues a year, indefinitely. The only thing you need to keep it running is still being alive and still building. Both of which are probably true for the next forty years.

Why it fits you specifically: you already write this. Your journal on alexisbouchez.com is, structurally, exactly this newsletter. You would just be committing to a weekly export. One edit pass per Sunday. Publish. Done.

Monetization honesty: paid tier, eventually. But Le Carnet is mostly a compounding reputation asset, not a revenue line. Ten years of this and you are the French indie developer everyone else has been reading in the background. Year fifteen, you sell a book version. Year twenty, you sell a bigger book version.

What would kill it: trying to make every issue "good." One page. Every Sunday. Bad issues are fine. The archive is the thing; no single issue matters.

9. Now the ones I left out of the first list

The filter I applied to the first seven was strict. I cut anything that touched your documented burnout triggers: SaaS, hollow market chasing, cold outreach at scale, fake enthusiasm. You told me to put them back. Fair. Here they are, with the triggers named out loud so we can see if each one actually avoids them or just reintroduces them in a prettier wrapper.

Your burnout triggers, on the record:

  1. B2B SaaS you do not use yourself. On March 21 you talked about killing Palmframe because it felt boring. You called it hollow.
  2. Cold outreach at scale. Resend suspended you in March. The outreach was not the problem; the volume was.
  3. Recording free product reviews. On April 7 you said you were "tired of recording free reviews."
  4. Building for hypothetical market needs. You explicitly reject "what the market supposedly needs."
  5. Fake enthusiasm. Once you notice it, you cannot push through it.

The format below is the same for every idea: a newsletter from the first seven, paired with the smallest possible SaaS, where the SaaS is something you would use yourself before anyone ever paid for it. The newsletter is distribution; the SaaS is monetization; the entry point for both is your own daily use. If you would not open the tool three times a week in your own browser, the idea does not belong here.

8. Infra Hebdo + InfraIndex

Idea 4, paired with the smallest possible SaaS: a search index over your own archive. Every infra item you have ever curated, tagged, searchable in one box. Free tier for the current month; 5€ per month for full archive search; 15€ per month for saved searches and alerts on new matches.

Burnout trigger check. Is the tool something you use? Yes. You curate hundreds of items a year; finding the one from eight months ago is a real problem you already have. Cold outreach? No, the newsletter is the distribution. Hypothetical market? No, the first user is you. What would kill it: user generated content, comments, "community" features. It must stay read only.

9. Oubliés + L'Encyclopédie Souterraine

Same model applied to the forgotten thinkers newsletter. Every weekly profile becomes a permanent entry in a searchable, cross referenced encyclopedia. Free to read the newsletter; 5€ per month for searchable archive, bibliography export, and the cross reference graph.

The cross reference graph is the actual asset. As you add thinkers over years you naturally build a web of influences: X read Y, Y translated Z, Z cited W. After three years nobody else has this. After ten, academic papers cite it. After twenty, it is a primary source in its own right.

Burnout trigger check. Do you use it yourself? Yes. You already have a de facto version in your research folder; it just is not searchable by anyone else, and not efficiently searchable by you either. Cold outreach? No, academics find it via search. Hollow? No, it is closer to a museum than a product. What would kill it: letting readers submit entries. You are the curator. That is the whole moat.

10. Le Carnet + TinyPress

The weekly journal newsletter paired with the smallest possible publishing tool: drop a Markdown file into a folder, it becomes an email plus an archive page, that is the entire feature set. No block editor. No AI assistant. No analytics dashboard beyond "number of subscribers." Closer to Apache httpd than to Substack.

You already run this infrastructure yourself on alexisbouchez.com. Wrapping it so other developers can use it is a few weeks of work. Charging 10€ per month for it is the smallest possible SaaS you could ship without lying to yourself.

Burnout trigger check. Would you use it? You already do, in a homegrown version. Cold outreach? No, Le Carnet itself is the distribution channel. Hypothetical market? The market is every developer who hates Substack's weight but does not want to roll their own Next.js site. That market is visible in every Hacker News thread on publishing. What would kill it: adding "collaborative editing" or a mobile app. If it is not ultra minimal, it is dead on arrival.

11. The Artisan Developer + The Craft Directory

Idea 2, paired with a directory of solo developers who have built one tool for five plus years. Verified, reachable, public profiles. Free to browse; 5€ per month to unlock contact info; sponsors pay 300€ to be listed as a craft company that is hiring.

You would use this tool yourself. You keep asking who to obsess over for 30 years, and a searchable directory of actual living craft developers is the exact reading list you want in your own browser on a Sunday afternoon.

Burnout trigger check. Do you use it yourself? Yes. Cold outreach? Only the newsletter interviews, which are not scaled cold email; they are one per week, warmed via shared audience. Hollow? No, it is a reference, not a pitch. What would kill it: unverified profiles. Curation is the entire product.

10. Fighting the giants: head to head with existing references

You asked me specifically about Technovelgy, startups.rip, and old school software archives. These are established niche encyclopedias with real audiences and real SEO moats. "Fighting" them is not about replacing them; it is about picking the axis they do not cover and owning that axis permanently.

The rule when fighting a giant: do not copy their format. Pick the axis they ignore. Different language. Different primary source. Different time cut. Different structural unit. A clone dies quickly. A well chosen complement lives next to the giant for decades and never gets dislodged.

12. Against Technovelgy: Fictives, une invention par livre, en français

Technovelgy.com is organized by gadget. One fictional invention at a time, with a book reference attached as metadata. Hundreds of thousands of visits a month. Solo operator for twenty plus years. Serious SEO moat in English.

The axes Technovelgy does not cover:

  1. French language. There is no French speculative tech reference anywhere on the internet. Jules Verne alone has more than 300 inventions nobody has catalogued this way in French.
  2. Book centric, not gadget centric. Each issue is one book, with its complete invention list, cross referenced to real tech as of the date of your issue. Technovelgy is a dictionary; you would be writing a bibliography.
  3. Closing the loop. Technovelgy rarely updates whether an invention actually got built. Your newsletter does exactly this: "this week, Verne's electric submarine: here are three real 2026 startups actually building it."

Arithmetic: the French speculative canon alone has hundreds of relevant books. Verne, Rosny aîné, Barjavel, Werber, Bordage, Damasio, Volodine, Houellebecq's speculative bits, Bellanger. Add the translations you read. You have a decade of primary material before you even start repeating yourself.

Monetization: Patreon for the newsletter, display ads on the archive, sponsor slots for "real startups building this now" features (French deep tech VCs would pay for exposure to that audience). Year five: co publish with a French SF press like Le Bélial' or Mnémos.

Why this beats Technovelgy specifically: you are not competing for English SEO. You are entering an empty room.

13. Against startups.rip: La Morgue Indie

startups.rip catalogs dead startups, mostly the VC funded kind. The format is a tombstone plus a one liner. It is brutal and short by design.

The axes startups.rip does not cover:

  1. Bootstrap and indie only. No VC corpses. Every dead startup in your graveyard was solo or small team, self funded, closed without raising.
  2. Founder interviews. Every entry has a real conversation, not just a tombstone. The founder's own post mortem in their own words. startups.rip cannot do this at scale; you can, at one per week.
  3. The counterweight. You already have a research file called dumb-ideas-that-worked-analysis. Pair La Morgue Indie with a monthly sister issue called Idées bêtes qui ont marché: one dumb product that somehow worked, per month. Readers of the morgue need the antidote.

Arithmetic: there are thousands of dead indie SaaS from the last decade alone. One per week is 520 post mortems in ten years. That is a reference the VC press cannot write, because it requires trust from indie founders, and indie founders only trust other indie founders.

Monetization: 10€ per month for the full archive. Sponsored by tools that serve indie devs (Stripe, Cal.com, Gumroad, Lemon Squeezy). Year five: a book called something close to The Indie Morgue.

Why this beats startups.rip specifically: startups.rip is a list. You would be writing the anthropology.

Burnout trigger note: you have said you are tired of reviews and cold outreach. This is not those. Interviewing dead founders is closer to oral history than to marketing. Almost nobody you interview will still be actively "in the game"; the emotional register is completely different from hustling a living product. That is why it does not re trigger the exhaustion you got from the review and affiliate hustles.

14. Against Folklore.org: Folklore Minitel

Folklore.org is Andy Hertzfeld's collection of stories about the early Macintosh. It is one of the warmest corners of the internet. It is also silent about every other computing tradition on earth.

The axes Folklore.org does not cover:

  1. Minitel. The entire French videotex ecosystem of the 1980s and 1990s is undocumented in narrative form. You have minitel-startups-revival-analysis in your research folder already; you already care about this.
  2. First person stories from the people who built Minitel apps, the first French ISPs, the Bull era, 3615 services, the earliest French BBS operators. Most of them are now in their 60s and 70s. Every year one of them dies and the stories die with them.
  3. French language. No Folklore.org equivalent exists in French. This hole will never be filled again if nobody starts now.

Format: one story per week. Interview, transcribed, lightly edited, one photograph, one technical artifact (a line of source code, a screenshot of a Minitel screen, a scanned receipt from a 3615 bill). The newsletter is the first draft of the eventual book.

Arithmetic: the French pre web computing world has thousands of stories. You have roughly a ten year window before most of the primary sources are physically gone. This is the newsletter with the highest urgency on the entire list.

Monetization: paid archive, book deal, and very plausibly a museum partnership. The Musée des Arts et Métiers or the Cnam would take this exact archive.

Why this beats Folklore.org specifically: Folklore.org is complete in English; Minitel is empty in French. You are not fighting for the same readers. You are recovering an era while its witnesses are still alive. That alone makes this the single most defensible idea on the list.

15. The Source Code Archive

One famous old source code reading per week. Unix v6. The first TeX. Fabrice Bellard's tcc. Tetris 1984 in Pascal. The original Doom release. SQLite at version 1.0. The first few commits of Ruby on Rails. The first redis.c, the first curl.c, the first nginx module. You read it, you extract one beautiful function, you explain what makes it beautiful, you post the full file in the archive.

Arithmetic: there are thousands of historically important codebases that nobody is close reading in a structured, weekly, archival way. Forty years of this and you have the definitive guide to "what good code looked like before AI wrote most of it."

Monetization: book deal (Stripe Press would buy this almost on sight), paid archive, university course licensing, sponsor slots from devtools companies who want to be seen next to craft. This is the most obviously fundable of the new eight.

Why this fights Technovelgy and Folklore.org at once: they tell stories about old technology. You would be reading the actual source of old technology. Different primary source entirely.

This is also the one that scratches your March 21 itch most directly. You asked which codebase you should obsess over for 30 years. Reading one famous codebase carefully per week for ten years is the cheapest possible way to find the answer. It is an apprenticeship that pays you while you take it.

11. Honest ranking, with the widened set

The first ranking I wrote only covered the seven purist newsletters. The eight new ones change the math, mostly in their favor: some come with built in monetization, and one (Folklore Minitel) has physical urgency. Here is the reranking, as tiers.

Tier 1: would launch these Monday

  1. Folklore Minitel (14). Highest urgency on the list. The primary sources are dying. Nobody else is going to do this. Ten year window, and every year you wait is a chapter that stops being possible.
  2. Oubliés + L'Encyclopédie Souterraine (9). Your deepest documented obsession, with a small SaaS built from what you would use yourself from day one. The cross reference graph alone is a twenty year asset.
  3. Le Carnet + TinyPress (10). Lowest friction on the whole list. You already write the content; you already run the infrastructure. The SaaS wraps what you already have. Starting is a formatting decision, not a new project.

Tier 2: strong, just slower or requiring more patience

  1. The Source Code Archive (15). Most fundable. Teaches you the craft while you build the audience. The cheapest apprenticeship in computing that also happens to be a publishable archive.
  2. The Artisan Developer + Directory (11). The SaaS paired version of idea 2; strictly better monetization without losing the soul.
  3. Fictives, une invention par livre (12). Empty French room, unique axis, clear SEO long game. Needs patience while you build the audience from zero.
  4. Infra Hebdo + InfraIndex (8). Shortest revenue path. The real risk is boredom in year ten, which is why I would only launch it if you already have a tier 1 idea running beside it to keep your enthusiasm high.

Tier 3: launch as second or third, not first

  1. L'Hérésiarque (5). Deep alignment; small initial audience. Best as a sister publication to Oubliés.
  2. La Morgue Indie (13). Strong format; slight risk of re exposing you to the "review hustle" feeling. Run it only once the tier 1 ideas are stable and your emotional relationship with indie SaaS has cooled.
  3. Codex (6). Most durable, slowest to reward. The research project of last resort.
  4. Katabase (3). Most beautiful, hardest to keep honest, highest risk of drifting into essayism. Probably the second book you publish, not the first newsletter you launch.

Tier 4: the purist versions you already read above

Ideas 1, 2, 4, and 7 in their original "no SaaS" form are still good. I am leaving them at the bottom of this reranking not because they are worse, but because their SaaS paired versions (8, 9, 10, 11) are strictly better on monetization without losing the core. Pick the purist version only if adding a SaaS would contaminate the enthusiasm for you; you will know if it does.

12. What I would actually do if I were you (updated)

Launch two. Not seven, not fifteen. Two.

Pick one from tier 1 that you can start this week, and one from tier 1 that you care about emotionally enough to keep writing on a bad Sunday. Let me name the two most likely pairings.

Pairing A (safe, high compound, low risk). Le Carnet + TinyPress, plus Oubliés + L'Encyclopédie Souterraine. Both are formatting decisions, not new projects. Both have a small honest SaaS attached. Both are powered by things you already do every day anyway. You could run this pair for forty years without friction. The downside: no single chapter of the story feels urgent.

Pairing B (urgent, historical, one real deadline). Le Carnet + TinyPress, plus Folklore Minitel. Le Carnet keeps the compounding reputation engine running with near zero effort. Folklore Minitel gives you a mission with a real ten year deadline: recover an era while its witnesses are still alive. Year twenty, you publish the book. The French tech establishment treats you as the custodian of a piece of its memory. You are on Radio France within five years.

The decision comes down to one question. Would you rather, in year five, be the person who built the definitive French language encyclopedia of forgotten thinkers, or the person who talked to the last surviving Minitel engineers before they died? Pick the one that makes you slightly uncomfortable when you imagine not doing it. That is the tell.

Wait at least a full year before adding a third. That feels slow. It is slow. That is the point. A newsletter you can keep running for forty years is, by definition, one you do not launch in a weekend and abandon in a quarter. The whole game is patience that compounds.

And then, when you are sitting at your desk in 2066, adding issue 2,080 of whatever you picked on a Sunday morning, you can check back here and see if this note held up. I hope it does.