17,652 indie launches analysed in May 2026. Vercel hosts a third of them. This month we look at what the other two-thirds are running.
Strip Vercel out of the May data and the indie web underneath looks nothing like the headline picture.
Last month we described a "modern indie recipe": Tailwind, React, Next.js, Vercel, Cloudflare, shadcn/ui, Google Fonts, Resend, Google Analytics. That recipe sits on top of a launch market where Vercel hosts about a third of every site.
Tailwind adoption drops from 54% to 46%. React drops from 36% to 20%. WordPress and jQuery, invisible in the headline ranking, surface as a real slice of the non-Vercel web. And Google Analytics, counterintuitively, gains share.
The modern-indie-recipe framing from April was mostly a description of what people deploy to Vercel.
Five things stood out
- Strip Vercel out and the stack changes shape. Tailwind 54% → 46%, React 36% → 20%, WordPress 6% → 8%, jQuery 1.8% → 2.5%.
- The modern analytics tools cluster on Vercel. Outside Vercel, Google Analytics is more common than in the Vercel-only slice: 43.3% of non-Vercel launches run GA, versus roughly a third of Vercel-hosted launches. PostHog and Plausible are concentrated on the Next.js-and-Vercel slice.
- Cloudflare Email Sending kept climbing in May. From 38 launches in April to 77 in May. Still a small base, but the bet we flagged in April is starting to pay off.
- Tuesday's dominance looks self-fulfilling. On platforms where founders pick the day, Tuesday peaks. On PeerPush, where most founders take a scheduled queue slot rather than choosing a day, the peak vanishes and Wednesday leads instead.
- 29 founders claimed their own launch and recrawled across April and May. 19 shipped at least one fix: llms.txt, a security header, a privacy policy, the easily-forgotten stuff.
A quick word on methodology
The scored launches come from Product Hunt (84.5%), PeerPush (9.1%), and Hacker News Show HN (6.4%). Every launch in this set was crawled during May 2026. Same baseline crawl as last month: static HTML, rendered DOM via Playwright, DNS and RDAP, legal-page discovery. Tech detection runs against 4,407 fingerprints. Full methodology at /methodology.
Two things changed since the April edition. First, two detection bugs in the April pipeline were fixed before May ran: a wildcard-DKIM bug that had inflated every ESP's count (patched 2026-05-08), and a DNS resolver that had been dropping about 15% of lookups (replaced 2026-05-21) and undercounting every DNS-derived signal. May runs on the fixed pipeline from day one, so its numbers are comparable to the corrected April figures, not the originally-published ones. There's more on both fixes near the end. Second, we now have two months of comparable data, so this issue introduces a delta table further down.
The cohort
| Source | Launches | Share |
|---|---|---|
| Product Hunt | 14,913 | 84.5% |
| PeerPush | 1,600 | 9.1% |
| Hacker News (Show HN) | 1,132 | 6.4% |
| Manual / other | 7 | <0.1% |
| Total scored | 17,652 | 100% |
Product Hunt stayed the dominant source at 84.5% of scored launches, up slightly from April's 82.1%. PeerPush slipped from 11.8% to 9.1%, and Show HN held around 6%. Product Hunt surfaces far more launches we can crawl than the other two boards combined, and that shape looks stable across both months.
The non-Vercel cut
A reader of the April edition asked on Reddit a question that turned out to matter more than it first looked: what do the stats look like if Vercel wasn't in the game?
We ran the cut against the 11,996 May launches that aren't hosted on Vercel. Reorder the chart and you'd expect the same technologies in a new sequence. That's not what happens; the composition changes. WordPress, which doesn't crack the top 15 in either view, sits just underneath at 8% of the non-Vercel cohort, up from 6% across the full one. The post's title is what it is for a reason.
The stack, with Vercel removed
| Technology | Full cohort | Non-Vercel cohort | Δ |
|---|---|---|---|
| Tailwind CSS | 54% | 46% | -8pp |
| React | 36% | 20% | -16pp |
| Next.js | 34% | 18% | -16pp |
| Vite | 24% | 27% | +3pp |
| WordPress | 6% | 8% | +2pp |
| jQuery | 1.8% | 2.5% | +0.7pp |
Tailwind's 8-point drop outside Vercel is real but modest; React and Next.js each lose closer to half their share, down about 16 points each. Vite picks up some of the slack: that's the "modern stack but not on Next.js" crowd, usually deployed to Railway, Render, or Netlify.
The bigger surprise is underneath the visible ranking. WordPress holds 8% of the non-Vercel cohort, up from 6% across the full one; jQuery shifts the same way, 1.8% to 2.5%. Neither cracks the top 15 in either view, so neither appears in the chart above. The tell is the table just before it: both rise once Vercel comes out. The slice of indie launches that doesn't deploy to Vercel still contains a real WordPress-and-jQuery population, which the modern-indie-recipe framing from April quietly hides.
Hosting, with Vercel removed
| Host | Share of non-Vercel cohort |
|---|---|
| Hostinger | 8.5% |
| Netlify | 6.8% |
| GitHub Pages | 5.1% |
| Railway | 3.5% |
| Google Cloud | 3.2% |
| Render | 3.0% |
| AWS S3 | 2.2% |
| OpenNext | 1.9% |
| IONOS | 1.5% |
| Plesk | 0.7% |
Hosting outside Vercel is genuinely fragmented. No single provider runs a runaway share. The "hosting monoculture" framing from April was almost entirely a statement about Vercel; the rest of the market doesn't have a runaway leader.
Analytics, with Vercel removed (and a counterintuitive note)
The counterintuitive bit: Google Analytics rises when Vercel is removed. Google Analytics is more common outside the Vercel cohort than inside it: 43.3% of non-Vercel launches run GA, against roughly a third of Vercel-hosted ones. That gap is a split between two halves of the same month, not a month-over-month wobble, so the roughly ten-point difference is structural, not noise. The modern product-analytics tools are the ones that cluster on Vercel: PostHog runs about twice as often inside the Vercel cohort as outside it, and Plausible leans the same way. The rest of the indie market still reaches for GA by default.
| Provider | Share of non-Vercel cohort |
|---|---|
| Google Analytics | 43.3% |
| Cloudflare Web Analytics | 8.1% |
| Microsoft Clarity | 6.5% |
| PostHog | 4.9% |
| Plausible | 3.7% |
| Umami | 1.4% |
| Sourcebuster | 0.8% |
| Hotjar | 0.7% |
The April note that "PostHog and Plausible lead the early data here but not at Resend-level margins yet" needs a caveat: PostHog and Plausible lead the Vercel-cohort data. The wider indie launch market still defaults to GA.
Email check: Resend still wins, even without Vercel
We didn't run "without Resend" because Resend's lead is already the headline. The question worth checking is the opposite: does that dominance survive when the Vercel cohort is removed?
It does. Resend still leads transactional email in the non-Vercel cohort at 13.8%, well above SendGrid, Mailgun, or any other provider in the same slice. So the "Resend ate the legacy market" finding from April is not a Vercel-cohort artifact. It holds across both halves of the market we can split.
Why the cut matters
The frame that matters for the rest of this post: there are two indie webs and they look quite different. There's the Vercel cohort, which fits the modern-indie-recipe description almost exactly. And there's everything else, which still has a meaningful WordPress and jQuery population, reaches for GA more than any other analytics provider, and picks its host from a wide, fragmented field. Charts that average across both halves implicitly weight the answer towards whichever cohort is bigger that month, and Vercel's third of the market does most of that weighting.
We'll keep running the "without Vercel" cut every issue from here, alongside the full-cohort tables. It's the same underlying data; the non-Vercel half is just the part April's headline averaged away.
Risers and fallers
Two months of comparable data, so for the first time we can show what's moving month over month.
| Field | April 2026 | May 2026 | Δ |
|---|---|---|---|
| Vercel hosting share | 33.2% | 32.1% | -1.1% |
| Tailwind CSS adoption | 58.0% | 53.9% | -4.1% |
| React adoption | 36.9% | 35.6% | -1.3% |
| Next.js adoption | 35.5% | 33.7% | -1.8% |
| Resend share of email | 20.4% | 18.5% | -1.9% |
| Cloudflare Email Sending | 0.22% | 0.44% | +0.22% |
| Named AI-builder cohort | 8.1% | 8.5% | +0.4% |
| Lovable share of named cohort | 53.4% | 51.9% | -1.5% |
| Avg Vibe Score (PH) | 31.9 | 32.1 | +0.2 |
| Avg Launch Readiness (PeerPush) | 90.6 | 90.7 | +0.1 |
| Avg Vibe Score (HN Show HN) | 20.5 | 20.0 | -0.5 |
For the adoption and share rows, anything inside about two percentage points month over month is within the noise band. A handful of lines are worth pulling out anyway, some because they moved more than that, some for reasons other than size.
Cloudflare Email Sending: 38 to 77 launches. The bet flagged in April is paying off, but modestly. Cloudflare Email Sending went from 38 launches in April to 77 in May. (The April post originally read 10 here; that was a DNS-resolver undercount, corrected to 38 along with the rest of the April email table.) The absolute number is still small enough that it doesn't appear on the visible top-10 of the email breakdown on our trends page. But the direction is right, and given the product is still in public beta with low awareness, steady growth off a small base is the pattern you'd expect before any steeper ramp. We'll keep tracking it. There's a plausible mechanism for a later jump: if CF Email Sending shows up in more code that future models train on, assistants may start defaulting to it. But at 77 launches, that's a hypothesis to test, not a trend to extrapolate.
Worth surfacing alongside this: Cloudflare Email Routing sat on 1,619 launches in May, a different product (inbound forwarding, not outbound sending) that has been around for years. It's the largest Cloudflare-in-email footprint in our data by a wide margin. The strategic read: Cloudflare already owns a meaningful slice of indie-domain inbound email handling, which is exactly the distribution surface that makes the outbound Sending bet less far-fetched than the April footnote framed it. Founders who picked CF Email Routing to handle their incoming mail are a natural conversion pool for CF Email Sending when they outgrow forwarding.
Lovable held its lead but stopped pulling away. It was 53.4% of the named-builder cohort in April and 51.9% in May, about half of every named-builder launch both months. As a share of all launches it's flat too: 4.3% to 4.4%. The chasers are still chasing. Replit (239 launches), Base44 (165 launches), Bolt (122 launches), v0 by Vercel (76 launches) are the next four, none close. No new entrant broke into the top tier.
Tailwind slipped most: 58.0% to 53.9%, down about four points. It's the only piece of the April recipe that moved more than the noise band, and it points back at this issue's spine: Tailwind is meaningfully more common in the Vercel-shaped indie web than outside it. Whether May is the start of a real drift or just a one-month mix change is one to watch.
Resend's share dipped from 20.4% to 18.5% too, measured on the fixed pipeline both months. That one is a real, modest decline.
The rest of the recipe barely moved: Vercel, React, and Next.js each held within about two points. The stack at the top of the cohort is sticky on a one-month window. Six months from now, the same stability may not hold; for May it does.
Tuesday is launch day, mostly
The folklore in indie circles is that Tuesday is the day to launch. We checked it against every launch in our corpus, not just May. The folklore is partly right. Tuesday does lead overall, but the platform split is where it gets interesting.
| Day | Launches | Share |
|---|---|---|
| Sunday | 4,170 | 9.9% |
| Monday | 5,440 | 12.9% |
| Tuesday | 8,023 | 19.1% |
| Wednesday | 7,277 | 17.3% |
| Thursday | 6,500 | 15.4% |
| Friday | 6,120 | 14.5% |
| Saturday | 4,571 | 10.9% |
All 42,101 launches we've crawled, by day of week (launched_at, UTC). This is the full corpus including earlier launches, not the May cohort.
Tuesday is the busiest day, and across the whole corpus it leads clearly: 19.1% of all launches against Wednesday's 17.3%. The spread among weekdays is small next to the gap to the weekend, though. Saturday and Sunday each run at roughly half Tuesday's volume, while the five weekdays sit in a 13% to 19.1% band. The takeaway isn't "launch on Tuesday." It's "launch on a weekday." The weekend is the only real cliff. The more interesting split is by platform.
By platform: the Tuesday peak only exists where founders pick the day
The aggregate table flattens out a real platform-shaped story. Splitting by source:
| Day | Product Hunt | PeerPush | Show HN |
|---|---|---|---|
| Sunday | 3,361 | 538 | 262 |
| Monday | 4,515 | 576 | 347 |
| Tuesday | 7,051 | 544 | 428 |
| Wednesday | 6,276 | 594 | 406 |
| Thursday | 5,529 | 584 | 387 |
| Friday | 5,244 | 575 | 301 |
| Saturday | 3,765 | 561 | 245 |
Product Hunt, PeerPush, and Show HN only. A handful of manually-added launches sit outside these three, so the columns don't sum to the corpus total above.
Product Hunt peaks on Tuesday with a steep weekend dropoff. Show HN follows the same shape on much smaller volume, peaking Tuesday. PeerPush goes its own way entirely: Wednesday is the busiest day on PeerPush, and the rest of the week is essentially flat.
The PeerPush difference isn't a founder-preference difference. On PeerPush, most founders don't pick a launch day. The default is to submit and join a queue, and the platform schedules the launch; you can pay to launch immediately, but we assume most don't. So PeerPush is the closest thing we have to a natural control: day-of-week selection pressure is mostly absent on the founder side, not amplified by folklore the way it is on Product Hunt.
So the Product Hunt Tuesday peak is probably self-fulfilling: founders launch Tuesday because the folklore says to, which manufactures the crowding the folklore predicts. PeerPush's flat distribution is closer to what indies look like when most founders aren't steering for the calendar.
So the practical question for a founder isn't "is the folklore right" but "do I want to be in the most-crowded slot of the most-crowded day, or somewhere else?" On Product Hunt, Tuesday and Wednesday are where you're competing with the most other launches for the same eyeballs. The contrarian play isn't the weekend, where you swap that competition for losing half the audience. It's the back of the week: Thursday and Friday are still weekdays but carry less of the Tuesday-Wednesday crush.
The claim funnel, two months in
In April we wrote up one anonymous case study: a founder who claimed their listing the day after launch and went from 2.0 to 8.5 in 31 hours. We treated it as a one-off. Two months in, it's a small but real cohort.
29 founders claimed their listing and triggered a recrawl between 2026-04-01 and 2026-05-31. Of the 28 we could compare before and after (one claim had no clean launch-day baseline to measure against), 19 shipped at least one observable fix; the other 9 were already in good enough shape that nothing changed on the recrawl.
That split is the story. Most founders who claim are already polished: the median claimer launched at a StackScope score of 7.3 and a Launch Readiness of 100. Claiming closes the last small gaps, it doesn't rescue a broken launch. The cohort didn't move from broken to fixed; it moved from good to very good, 7.3 to 9.5.
| Launch-day baseline (median) | After claim + recrawl (median) | |
|---|---|---|
| StackScope score | 7.3 | 9.5 |
| Launch Readiness | 100 | 100 |
The most-added items across the claim cohort:
- llms.txt (11 added)
- security.txt (9 added)
- Referrer-Policy header (9 added)
- Privacy policy (6 added)
- robots.txt (3 added)
llms.txt and security.txt are good-practice files we report but don't fold into the StackScope score. The score movement here came from the other three: the security header, the privacy policy, and robots.txt.
None of these are hard. They're the boring last 10% of a launch that's easy to forget.
If you launched something on Product Hunt, PeerPush, or Show HN in May and haven't checked your listing yet, find it on /browse. Claiming is free.
New this month: if your site was never posted to a board we crawl, the Launch Readiness Check runs the same analysis on any URL. Private to you, and free.
We had to correct the April issue twice
The April edition now carries two correction notes, and both of them are mine.
The first was a fingerprint bug: a rare wildcard DNS record was making our DKIM probe over-count email providers. The second was bigger and took longer to find. The crawler's DNS resolver had been quietly dropping about 15% of its lookups for weeks, which undercounted every DNS-derived signal, Resend most of all. The "Resend is dominating" finding was real. The numbers under it were not.
I caught the second one because a trend looked wrong. Resend's share appeared to slide month over month, and that did not match anything else I knew about where the company was heading. Chasing that "that can't be right" feeling is what surfaced the resolver. The month-over-month tracking this publication exists to do is also, it turns out, what catches its own mistakes.
Here is the part I'd rather not write. Both corrections happened to make the headline finding stronger, not weaker. That is luck, not vindication. A measurement bug can invent a trend just as easily as it can hide one, and the next one might not break my way. So if a number in any edition looks wrong to you, it might be. Tell us, and we'll re-run it in public.
What's next
This is issue #2. The shape is settling: a feature angle that's specific to the month, plus a delta table that runs every issue, plus a methodology note that gets shorter each time as the long-form lives at /methodology. The "without Vercel" cut is also going to be a recurring feature from here.
The next issue, covering June data, is the hosting and infrastructure deep dive: which CDNs front the most launches, the long tail of hosts beyond the big platforms, and who runs the DNS underneath. The Vercel concentration finding in this issue is the starting point. If you operate hosting infrastructure and want to see how your customers' launches show up in our data, drop us a note before we start writing.
Bookmark /blog.
Credits and links
Thanks to the PeerPush team for the ongoing integration support.
The spine of this issue, the "without Vercel" cut, was prompted by a Reddit comment on the April post asking the question we hadn't thought to ask ourselves. The sharpest edits to this publication tend to come from readers in exactly that shape. Keep them coming.
Explore the May data
Claiming a listing and triggering a fresh crawl are free.
- Find your launch: search by name and see how your May launch was scored.
- Claim a listing: from your launch page, use the "Is this your site?" link to verify ownership and trigger a fresh crawl.
- Browse the tech catalogue: stackscope.dev/tech.
- Bookmark /blog for the next issue, a hosting and infrastructure deep dive.