Marc Lou vs beehiiv: The Vibe-Coded Newsletter Fight

Tyler Denk asked Marc Lou if he had missed anything.

The question sat under screenshots from a newsletter called “Built different.” In that issue the beehiiv CEO wrote that “A 92.9% delivery rate is rough,” set it against what he called beehiiv’s “industry-best 99.1%,” and described the audience as “a couple hundred basement dwellers who treat this guy’s Stripe dashboards like an OnlyFans account.” The excerpt post went up on 20 October 2026. A snapshot the next afternoon showed about 252,000 views.

Lou had started this, as a customer. He said beehiiv had taken a yearly bill from $1,000 to a future charge of $4,000. He vibe-coded a replacement, open-sourced it, and launched the paid version by quote-posting a feature list Denk had published on 14 October. That list did not say Lou’s name.

Pieter Levels had posted “Software is dead” on 11 October. He read the excerpt and said Denk sees Marc as a threat. The fortnight is a live test of that line: a payer became a competitor in days, and the company that sells the category answered with a real argument about email, then with insults.

View counts below are a snapshot from about 13:00 AEDT on 21 October 2026. Bills, list sizes, open rates, and delivery rates that a founder posted about their own product are self-reported. The exception is TrustMRR, which labels its MailCheap figure as Stripe-verified.

A customer, then a churn post

Marc Lou is @marclou. Tyler Denk is @denk_tweets, cofounder and CEO of beehiiv, and the creator of Big Desk Energy. The company account is @beehiiv. The longer public portfolio, the decade and the products people already know, is a different story. The short map is the Marc Lou page. This one starts with an invoice.

On 3 July 2025 he posted a recipe. “How to 6x your MRR in 1 minute.” The steps were “Triple the pricing,” “Reduce usage by 50%,” and “Add no extra value.” Then: “I want to leave @beehiiv now.”

He was still a customer more than a year later. On 3 October 2026 he posted “How to 12x your MRR in 1 minute”, then “@beehiiv keeps increasing its prices every year. As a customer, I feel ripped off. So I’m churning.” That post had about 336,000 views in the snapshot.

The bill he described on 9 October is the concrete version. He put the first yearly bill at $1,000, and wrote that the “future bill would have been $4,000.” Both figures are self-reported. MailCheap’s site puts the same complaint in his words: “I paid beehiiv for two years. They kept raising prices and didn’t add one feature I used.” Elsewhere on that site he says beehiiv “raised prices for the 3rd time in 2 years.”

April 2024 direct messages, which he posted on 20 October, already show a beehiiv site at marclou.beehiiv.com and a newsletter he described then as having “17K readers.” The site’s “two years” is the summary he wrote later. The posts checked for this piece do not show a beehiiv sponsorship or an affiliate deal.

On 5 October, before MailCheap existed, other people were already doing this arithmetic in public. Nathan Barry, who runs Kit, defended paying for sending infrastructure on 5 October: “spending 1.5% of that ($1,390) on high quality infrastructure is a rounding error.” He also said Kit is free up to 10,000 subscribers. Noah Kagan, of AppSumo and SendFox, posted the same day, “I’m confused why people pay so much for Kit and beehiiv,” and “Actually sending the emails: about $4/mo.” Those posts predate MailCheap. They are the price argument, already running.

The product he actually shipped

On 9 October he wrote “Bye bye Beehiiv.” “I vibe-coded my own newsletter platform.” He credited “a few prompts with Opus 5.5,” which is his name for the model. The stack he named was Amazon SES plus MongoDB, and he expected it to cost “less than $300/year.” He later put the volume at about 1.4 million emails a year, at $0.16 per 1,000 on the SES Essentials plan. That volume is self-reported. The post had about 400,000 views.

The next day the code was public. The first commit on marclou/mailcheap, at 12:36 UTC on 10 October, is titled “Mailcheap: self-hosted newsletters on Amazon SES.” It says mail goes out in batches through QStash. The repo was created at 12:55 UTC the same day. The license is AGPL-3.0. The language is JavaScript.

His announcement followed at 13:51 UTC. “MailCheap — Self-hosted newsletters on Amazon SES.” He priced that version at “$0.10 per 1,000 emails” and called it “92% cheaper than Beehiiv (for me).” About 370,000 views and 3,700 bookmarks in the snapshot.

The GitHub snapshot at the same hour on 21 October: 1,275 stars, 324 forks, 13 open issues, about 101 commits. The description reads “Newsletters where you pay for the emails you send, not for dormant subscribers.” Recent commit titles begin “Sync from mailcheap-cloud.” That wording suggests a private cloud repo mirrored into the public one. Treat the mirror as an inference from the titles, not as something he announced.

The hosted product is a different machine from the repo. mailcheap.ai headlines “Run your newsletter up to 83% cheaper than beehiiv, Mailchimp and Ghost.” The footer and FAQ say “Marc Lou built it, and JustShipIt Pte. Ltd., a company in Singapore, runs it.” Pricing on the site in October 2026 is per email sent, from $19 a month. The tiers listed are $19, $39, $79, $149, and $349. Starter is “$0.95 per 1,000 emails.” There is no free trial and no “Powered by” badge. The self-hosted SES pitch was a tenth of a dollar per thousand. The hosted pitch is $0.95. Those are different products with the same name.

On 15 October he said the hosted version sends through PostShiba, by Jesse Hanley, and “not AWS SES.” He added “not affiliated.” Hanley is a party to this fight, and a vendor. That conflict is disclosed again below, next to what he posted.

The site also states guardrails: a 30-day warm-up with throttled send sizes, a 3-hour test batch for new contacts, an auto-pause above 2% bounces or 0.08% spam complaints, a block above 5% or 0.3%, per-customer isolation, one-click unsubscribe, and DKIM, SPF, and DMARC. It offers an MCP and API connector for ChatGPT, Claude, Grok Bot, Claude Code, Cursor, and Codex, and it says agents have to ask before they send to the list.

How long it took depends which sentence you use. On 9 October the personal tool was “a few prompts.” On 20 October he agreed the product was more like “20–30 hours,” and wrote “I started working in it the day I released the open source repo.”

List size, also from the site, and self-reported: his own newsletter “with over 50,000 subscribers,” and Unicorne, TrustMRR’s newsletter, “with over 15,000.” Unicorne is at unicorne.trustmrr.com. The April 2024 message said 17,000 readers. Same founder, three windows.

A short timeline

Dates here are UTC, which is how the posts are stored. In Melbourne, on AEDT, anything after about 13:00 UTC falls on the next calendar morning. The 13 October send, the 14 October feature list, the 15 October launch, and the 20 October excerpt all did.

View counts are the 21 October 2026 snapshot, about 13:00 AEDT. Dollar figures in the 9 October row are self-reported.

A feature list with no name on it

On 13 October he posted, “I just sent my first newsletter issue with my vibe-coded Beehiiv alternative for 10% of the cost.” The screenshot was taken during the send: 576 of 2,000, on the Unicorne list. Delivered 92.9%. Opened 3.9%. Clicked 1.3%. About two hours later he posted a different open rate: “28% open rate on first issue blasted 2hr ago.” Hold both numbers. The newsletter on 20 October uses the first pair.

Denk’s viral post came the next day. Marc’s name is absent from it. “every week someone tells me they’re going to vibe-code their own @beehiiv alternative over the weekend.” Then a long list: 99.3% delivery, IP warming, suppression lists, bot-click filtering, WYSIWYG, an API, MCP, roles, a website builder, and more. The close is “but at least you saved $59/mo not paying for @beehiiv anymore.” @beehiiv reposted it. Snapshot: about 974,000 views, 3,400 likes, 1,090 replies, 546 quotes, 3,400 bookmarks.

Arvid Kahl quote-posted the list. “Hey Claude, implement every item on this list. Make no mistakes.”

The same night, into 15 October, Denk posted the counterexample from his own week. He had vibe-coded software for MADRE, his Medellín wellness club, instead of paying $79 a month for Mindbody. “‘building it’ is the easy part.” The build took “60+ hours,” now costs “$300 /mo,” and he is “on call 24/7.” That is a first-person bill for maintenance, from the person on the other side of this argument.

He also pointed at an older essay. On 15 October he linked “SaaSpocalypse,” from 1 September, his longer case against the idea that rented software is finished. In it he says the people making that case “probably think that Opus is some sort of peptide.” The October fight has a September draft.

On 19 October he wrote that the plan was “the next 6 weeks: relentlessly shipping as many high impact features and clearing the backlog.” Numbered feature threads followed. On the morning of 21 October he was still on them, at a mobile editor. Asked about price, he said “we feel great about our prices” and “you pay for what you get.”

He launched by quote-posting that list

MailCheap.ai went up on 15 October as a quote of the feature list. “I vibe-coded beehiiv, and it’s 83% cheaper.” “Introducing MailCheap.ai.” Open source, agentic, pay per email, “From $19/month,” no “Powered By” badge. Snapshot: about 1.27 million views, 3,700 likes, 521 replies, 3,100 bookmarks.

The next post is the comparison he wanted screenshotted. “Math: 4,000 subscribers, one email a week: $19 on MailCheap vs $109 on beehiiv Lite.” The site repeats that pair.

Steven, who writes a creator-tools newsletter, answered with the other end of a per-email price. A list of 50,000, sent daily, is “1.5M emails a month.” “That’s $349 on this which is literally more than beehiiv.” About 134,000 views. He and Lou then argued about Steven’s open rate. The posts checked here do not include a number for that rate.

Lou’s revenue post the same afternoon showed a screenshot at $77. “My new startup is profitable from day 1.” “I have no investors to please, so I won’t raise prices every few months.” On 18 October he called it “3-figure MRR now for MailCheap.AI,” and said delivery was “currently 99.2%, but there isn’t enough data yet.” On 19 October he shipped a help center, segments, a drag-and-drop editor, and eight themes, and wrote “Vibe coding software is my happy place.”

TrustMRR, in the 21 October snapshot, showed “$634 last 30 days” and “$634 MRR,” marked as verified Stripe revenue. That is the checked number. The $77 and the “3-figure” line are his earlier posts. mailcheap.ai/open said, for the 30 days to 20 October, a delivery rate of 99.6%, 50,398 emails delivered, and a spam-complaint rate of 0.008%. Those dashboard figures are self-reported, and the volume is small.

Levels reposted the launch the same hour. He did not add a long take.

On price, Lou’s line a few days later was “When software is becoming a commodity, raising prices is anti growth.” Nevo David, who makes Postiz, posted on 17 October that beehiiv “will charge me an extra $140 for the same plan. 42.5532% increase.” Lou replied, “Classic move. That’s why I left.” David’s product already has its own story, about agents as customers. Here he is simply another payer reading a renewal.

Asked about paid promos, Lou answered “Never.”

“Built different” went after him

The personal piece is the 20 October issue. The 14 October list left his name off.

“Just because you can build it doesn’t mean you should.” “A somewhat popular indie hacker recently decided to build his own beehiiv alternative.” Then the 92.9% line, against “industry-best 99.1%,” and the 3.9% open rate against a beehiiv average he put at “~41%.” One footnote belongs here. The 14 October list claimed 99.3% delivery. The 20 October issue says 99.1%. Both numbers are his, and they do not match.

The tone is the part the replies seized on. “a couple hundred basement dwellers.” “a circus of 38 bastard children masquerading as ‘startups’.” “their revenue charts all look like shit.” “The truth is — no one serious is ever going to use this.” He also wrote “We had a record week last week for both revenue and growth.” And: “Word to the wise: don’t conflate social media followers with competence.”

The issue is not only insults. The substantive case, in that same URL, is specialisation. He uses a division-of-labour argument, with Shopify and Thomas Thwaites’ toaster as the pictures. He lists the costs that keep arriving after the demo: SES, hosting, a database, auth, monitoring, validation, rendering, a CDN. He names Cloudflare Enterprise, three layers of email verification, and a click-verification vendor at “hundreds of thousands of dollars per year.” He puts beehiiv’s headcount in the piece at “130 specialists.”

The chart in the issue reads $2.91 million MRR. Multiply by twelve and you get about $35 million a year. That multiple is an inference from the screenshot, not a figure he typed. What he did type, the day before, was “sprinting towards $40M ARR by EOY,” plus “no one serious is moving to marc’s vibe coded project.” Travis Biziorek had called the feature sprint “fascinating watching what feels like panic in public from Tyler.” Denk called that a “pretty amateur take.” Biziorek’s post had about 121,000 views.

A separate public report sits beside those self-reported charts. Reuters, on 20 January 2026, said beehiiv expects 2026 revenue to “nearly double to $50 million,” with over 40,000 monthly active users and nearly 15,000 paid subscribers. TechCrunch, on 30 April 2024, reported a $33 million Series B led by NEA.

@beehiiv reposted the 14 October list and the feature-sprint post. No original statement from that account turned up in the posts checked.

Lou’s reply to the excerpt was one line, and it traveled. “Just marked your issue as ‘not spam’ to help your deliverability.” About 23,000 views and 519 likes. He also posted the April 2024 messages and wrote “8+ days to fix a bug,” under the line “we handle thousands of support tickets.” The screenshots show a Search Console problem on his beehiiv site. Denk replies “shouldn’t have” and “let me flag” on 12 April, then hands it to staff around 17 April. “8+ days” is Lou’s description of that gap.

He posted a friendlier message too, with “2 weeks ago btw.” The crop shows Denk writing “big fan of your work, don’t want you to feel slighted at all” and then congratulating him on HYROX and on September. The crop itself has no date. The “two weeks” is Lou’s caption.

The 92.9% figure was mid-send

The delivery attack uses a screenshot taken mid-send.

576 of 2,000 had gone out. Delivered 92.9%. Opened 3.9%. Clicked 1.3%. The list was Unicorne. Lou’s later post in that same thread says the open rate was 28% after two hours. On 18 October he said delivery was “currently 99.2%, but there isn’t enough data yet.” The public dashboard then showed 99.6% across 50,398 emails in the 30 days to 20 October, with spam complaints at 0.008%. All of those rates are self-reported. The dashboard volume is still a small sample next to beehiiv’s.

Denk’s 99.1% and his 99.3% are also his. So is the “~41%” open-rate average. A fair reading is that a half-finished send is a weak exhibit, and that a brand-new sender’s own dashboard is a weak exhibit too. Neither number, on this record, is an audit.

Hacker News barely picked the launch up: one submission, two points, no comments. Reddit was not available when this was checked, so Reddit’s reaction is unverified and left out.

Deliverability, price, and the work after launch

Strip the nicknames out and Denk’s essay matches what people who already run email businesses posted that week.

Jesse Hanley, who runs Bento and PostShiba, wrote that “@marclou can fail if he doesn’t take compliance, KYC, and abuse seriously.” He is not a neutral commentator. MailCheap’s hosted product sends through his PostShiba, which Lou disclosed on 15 October. On 20 October Hanley also wrote that “the Beehiiv business is unprofitable.” Read that one with the vendor relationship in view.

Thomas Crombie-Evans, of EmailOctopus, opened with “Running @emailoctopus the OG hosted SES wrapper” and “I have a few thoughts.” The rest of that thread was not read for this piece, so it is not paraphrased here. Geoff Roberts, of Outseta, was direct: “Building email for others is WAY more involved than you might think.”

The shared point is the gap between sending your own newsletter and sending everyone else’s. Abuse, complaints, and reputation sit on the same pipes. A self-hosted SES setup for one list, which is what the 9 October post describes, is a smaller job than a hosted product that takes strangers’ lists. Lou’s site lists warm-up, pauses, and authentication. The founders above are describing the years after the README.

The pricing objection is narrower, and it is real. Per-email billing is cheap for a weekly letter to a cleaned list, which is Lou’s example at $19 against $109. It gets expensive if you mail a large list every day, which is Steven’s $349 example. Lou’s counter is the repo description: you pay for the emails you send, “not for dormant subscribers.” That is a product choice. It is also a reason some senders will do the maths and stay.

Feature depth is the third solid point. Denk’s list is long on purpose. beehiiv sells an ad network, paid subscriptions, boosts, a website builder, team roles, and podcasts. MailCheap’s public pitch, as checked in October 2026, is sending, an editor, segments, a help center, themes, and an agent connector. A weekend clone of the send button does not arrive with the rest of the suite.

Maintenance is the fourth, and it is his own story as much as a warning. The MADRE build is 60-plus hours, $300 a month, and a founder on call, to replace a $79 tool. Yoshik put the general version in one sentence: “AI makes building software cheap, but your time is still expensive.” Launch Llama asked it outright: “When it comes to maintainability, do we really want to be maintaining all of these vibe coded projects .. ?”

Charts, a HYROX line, and the tone

“their revenue charts all look like shit” is a swing at the whole portfolio. The screenshots in the issue lead with small projects, including PoopUp at $9 a month. A second chart, in his follow-up, showed TrustMRR at $47,300 a month and DataFast at $31,900 a month. He captioned the set “here are his best ‘startups’” and “same trend.”

Lou answered “TrustMRR all time high last month” and “DataFast all time high last month”, and that the “Last datapoint is October and we’re the 20th.” A falling line on a chart that includes an unfinished month is a partial month. That part of his reply is a description of the x-axis.

ShipFast is a cleaner case, because Lou posted it himself the same day. “ShipFast in 2024: $126,000/mo” and “2026: $2,000/mo.” Those are self-reported. The drop is in his own words. CodeFast is on the charts Denk called the same trend. Lou’s reply does not give a CodeFast figure, so this page does not invent one.

The count of “38” is roughly his own bio. The @marclou profile lists seven named products and then “+30.”

The HYROX line in the issue says he “made more money last month selling ad space on his body during a HYROX race than from all of his software projects combined.” That is roughly what his own 30 September post shows, and it is a one-off month. He listed HYROX at $112,000 against TrustMRR $53,000, DataFast $31,000, CodeFast $7,000, Ship or Die $3,000, ShipFast $2,000, and smaller items. The software lines that are visible add up to something on the order of $97,000, before whatever the cut-off tail contained. The tail of the post was truncated, so an exact software total is unverified. The race money is also self-reported in that post. The closed auction itself, slot by slot, is the HYROX story. A sponsorship month can outrun a software month once. It does not make the software lines fictional, and it does not make the race a business model.

The tone is what turned a product argument into a pile-on. Aodhán Moran: “An ad hominem like this is a bit unusual coming from you.” Nicolas Le Roux: “What a disgusting way of communicating.” Denk’s reply to him was “I wouldn’t take painting advice from someone who’s never painted before.” Pablo: “The best reply to a clone is a better product, not insults.” Axel Schapmann: “seeing the CEO of a company that size repeatedly go after an indie hacker… feels a bit strange.”

The other direction exists too. Adrian Horning: “Serious companies I can promise you are not switching to Marc’s POS product.” Kane Kallaway wrote “guy’s a jester” and aimed elsewhere: “The actual threat to beehiiv is that humans will stop reading their email.”

Noah Kagan made a “Click each guy” game and linked a SendFox page. That page was not opened for this piece, so the game itself is undescribed here.

Levels picks a side

Levels was early, and he was brief.

On 1 October, before any of the MailCheap posts, Jon Yongfook asked what to use for a simple newsletter. Levels replied: “Vibe code your own then use Cloudflare Email Sending to send it on newsletter.* subdomain.”

On 11 October, two days after “Bye bye Beehiiv,” he posted the two lines that the rest of tech argued about for a week. “Software is dead.” “It’s now the era of personal software and everyone just builds what they need whenever they need it I think.” About 1.66 million views. The post is not about beehiiv. The calendar puts it in the middle of this story anyway. The replies, the moat list, and the API he priced afterward are their own piece. The cost version, him cancelling his own vendors, is the stack story. Whether that cheapness kills the job is the longer argument. His page keeps the years from collapsing into the slogan.

On 15 October he reposted MailCheap. On 20 October he quote-posted the excerpt. “He seems so incredibly triggered by @marclou.” “Living rent free in his head for a week now.” “So yes he sees Marc as a threat!” About 104,000 views. In the replies he added “It’s like when celebrities fight” and “Yep it’s very guerilla marketing, great work by him.” He did not discuss delivery rates, complaints, or compliance.

Denk’s answer to him was “I just want to see him succeed.”

Levels’ stance is clear. He treats the CEO’s reaction as evidence that the clone matters, and as free attention for the person who shipped it. That is a marketing read. It leaves the email-operations argument where Hanley, Roberts, and Denk actually put it.

What this week is a test of

One reading is the slogan. A customer who hated a renewal shipped a sender in public, under AGPL-3.0, and put a hosted version on sale at $19. Stripe-verified revenue on 21 October was $634 in the last 30 days. The incumbent CEO answered with a chart at $2.91 million MRR, a long feature list, and a newsletter with his name in it. If “software is dead” means a person can now build the thin version of a tool they are renting, this week happened.

The other reading is the bill that shows up after the demo. Shared email lives or dies on who else is sending, how fast you notice abuse, and whether anyone is awake when a domain’s reputation moves. Denk’s MADRE story, Hanley’s compliance warning, Roberts’ one-liner, and Steven’s daily-send maths are all versions of that bill. A 99.6% dashboard on about 50,000 self-reported emails does not retire it. A 92.9% frame from the first 576 sends does not settle it either.

Lou has a real grievance on price, documented in his own posts across 2025 and 2026, and a real product with a public repo, a company name, and a Stripe figure you can look up. Denk has a real case on deliverability, compliance, maintenance, feature depth, and the way per-email pricing behaves for people who send every day. The names he reached for, once the case was already made, handed Lou the screenshot.

A note on the numbers

The $1,000 and $4,000 bills, the “under $300 a year” SES estimate, the 1.4 million emails, the 50,000 and 15,000 subscriber counts, the $77 screenshot, the “3-figure MRR” line, the 99.2% and 99.6% delivery figures, the 28% open rate, and the $2.91 million MRR chart are self-reported by the person who benefits from the telling. TrustMRR’s $634 is the one MailCheap revenue figure in this piece that the site itself marks as Stripe-verified, and it is still a snapshot from 21 October 2026. GitHub stars, forks, and every view count are from about 13:00 AEDT that day. beehiiv’s 99.1% and 99.3% delivery claims disagree with each other. The HYROX comparison uses Lou’s 30 September list, with the tail of that post cut off. Reuters and TechCrunch are linked above for the company facts that are not a founder’s screenshot.

What do you think it means?

A price rise turned into a public competitor in under two weeks. An open-source repo and a $19 plan are now sitting next to a company that has raised a Series B and is posting seven-figure months. The hard part of email did not get vibe-coded away in the thread.

If you indie hack, the question is whether a tool you pay for is now a prompt plus a weekend, or whether the thing you pay for was never the screen. If you sell SaaS, the question is what is left once a customer can ship the thin version in public and the argument moves to reputation, support, and the features they still cannot be bothered to rebuild.

What do you think this fight means for indie hacking, and for SaaS?

FAQ