Aiolum Submit a story
All stories
AI film generator and editor that works from one sentence · United States

Koe

Solo founderB2CVideo
Outcome
US$88K a year run rate, October 2026Revenue falling
Launch
10 April 2026
Team
1 named founder, who writes solo founder; no headcount in any source
Outside funding
None found
Group
Video generation
Main channel
Launch sites and deal sites

Koe is a web app that turns one sentence into a short film. It writes a storyboard, a list of scenes with one still picture for each scene, keeps the same characters from scene to scene, and then sends each scene to a text-to-video model, an AI model that turns a description or a picture into a moving clip. Koe does not train these models. They are made by Lightricks (LTX), Kuaishou (Kling) and ByteDance (Seedance), and Koe pays for each use through an API, an interface that lets one program buy work from another company's servers. Users pay with credits, prepaid units that each generated scene uses up. Plans cost US$12 to US$600 a month.

Koe launched on Product Hunt on 10 April 2026 and got 6 upvotes. The launch was submitted by Hong Xuan Kow, and the maker's comment says "I'm Xuan, solo founder of Koe". On 14 April Koe added ByteDance's Seedance 2.0. Stripe records, read by TrustMRR, show US$645 of revenue in May, US$7,177 in June and US$32,697 in July. On 16 July the founder posted a Stripe screenshot with about US$10,100 in MRR, monthly recurring revenue, the subscription revenue of one month. Then revenue fell: US$20,414 in August and US$8,616 in September. On 14 October 2026 the MRR was US$7,352 from 360 subscriptions, a run rate of US$88,224 a year, and total revenue was US$73,304.

The revenue numbers are grade A, because TrustMRR reads them from Stripe, the payment processor, through an API key. The founder's posts and Koe's own pages are grade C. No source explains the fall after July, and no source gives a headcount. Under terms dated 15 September 2026, Koe also offers to turn authors' books into paid video series. The page also shows five earlier companies that promised easy video for everyone or sold stories as short paid video.

Koe: US$88K a year run rate, October 2026. AI film generator and editor that works from one sentence, United States, solo founder, evidence grade A.
The story in one card. Share it with the address of this page.
01

The numbers, graded

Each row shows the evidence grade and the source numbers from the list at the end of this page. The grades are explained in How Aiolum verifies.

A video app that made US$32,697 in its best month and had US$7,352 in MRR six months after launch, with revenue verified from Stripe
FactValueGradeHow we knowSources
Monthly recurring revenue, 14 October 2026US$7,352ATrustMRR, from the Stripe API. 360 active subscriptions. Two days earlier the profile showed US$7,568 and 367 subscriptions.[1]
Run rate from that MRRUS$88,224 a yearcalcCalculated: 7,352 times 12. Revenue in the last 30 days, US$7,380, is almost equal to the MRR, so one-time payments do not raise this figure.[1]
Verified revenue, all time to 14 October 2026US$73,304ATrustMRR, from Stripe. Equal to the last-12-months figure, so there was no revenue before April 2026.[1]
Revenue by month, 2026April US$70, May US$645, June US$7,177, July US$32,697, August US$20,414, September US$8,616, 1 to 14 October US$3,684ATrustMRR monthly timeline, in UTC calendar months.[1]
Best monthJuly 2026, US$32,697A4.56 times June, and 44.6% of all revenue to date (calculated).[1]
Fall from July to September 202673.6%calcCalculated: 1 minus 8,616 divided by 32,697. August was 37.6% below July, and September 57.8% below August.[1]
Last 30 days against the two months beforeUS$7,380 against US$16,677 a month, or 44%calcCalculated: 40,734 for the last 3 months minus 7,380 for the last 30 days, divided by 2, is 16,677; 7,380 divided by 16,677 is 0.443. TrustMRR shows the last 30 days 22.5% below the 30 days before them.[1]
MRR, 16 July 2026about US$10,100CFounder's post: Koe has reached 10k in MRR, and monthly revenue grew 10 times in less than a month. The attached Stripe screenshot shows US$10.1K MRR, US$32,200 gross volume and US$30,400 net volume from sales for April 2025 to July 2026.[2]
Change in MRR, 16 July to 14 October 2026minus 27.2%calcCalculated: 7,352 divided by 10,100, minus 1. Revenue fell much faster than MRR.[1, 2]
July revenue against the mid-July MRR3.24 timescalcCalculated: 32,697 divided by 10,100. The difference comes from payments outside the mid-July MRR: one-time credit packs, yearly plans paid in advance, and subscriptions that started after 16 July or ended soon after they started. The sources do not split it.[1, 2]
MRR per active subscriptionabout US$20.42 a monthcalcCalculated: 7,352 divided by 360. Between the monthly prices of Hobby (US$12) and Starter (US$30).[1, 7]
Profit margin, last 30 days93.0%CA TrustMRR snapshot that is not in its list of metrics read from the payment provider. The profile does not say which costs it subtracts, so this page does not use it.[1]
Launch10 April 2026, on Product HuntC6 upvotes, 1 comment, 68th product of the day. The first Stripe revenue, US$70, came in the same month.[3, 4, 1]
Founded date on the TrustMRR profile8 April 2025CEntered by the founder. The Stripe screenshot also starts its range in April 2025, but revenue starts in April 2026.[1, 2]
Team1 named founder; no headcount in any sourceCTrustMRR and Product Hunt name one founder, who writes 'solo founder of Koe'. The careers page says 'We keep the team small' and lists open roles in research and engineering.[1, 4, 13]
Outside fundingnone foundCKoe is not in the Y Combinator company directory, and no source reports a funding round. The About page gives an email address for press and investors.[15, 12]
Video models on the plansLTX 2.3 Pro, Kling 3.0 and Kling O3, Seedance 2.0CPricing page feature table. The privacy policy names ByteDance/BytePlus (Seedance), OpenAI, Anthropic, Google, fal.ai, Lightricks/LTX and ElevenLabs as AI model providers.[7, 9]
Credits for one 6-second scene on the entry modelabout 60 for motion plus 20 for the start frameCPricing page answer for LTX 2.3 Pro. It adds that Kling and Seedance cost more, without numbers.[7]
What a user pays for that sceneUS$0.80 to US$4.79, or US$0.13 to US$0.80 a secondcalcCalculated: 80 credits at US$0.01 a credit (Ultra, US$600 for 60,000 credits) up to US$0.0599 a credit (a pack of 100 credits for US$5.99). On Hobby, at US$0.015 a credit, the scene costs US$1.20.[7]
List price of a 6-second clip at model providersUS$0.48 to US$2.40calcCalculated: 6 seconds of Veo 3.1 on the Gemini API at US$0.08 a second (Lite, 1080p) to US$0.40 (Standard). Kling v3 Pro on fal.ai is US$0.14 a second, or US$0.84 for 6 seconds. Koe's own costs are not published.[16, 17]
Time from launch to the October 2026 figure6 monthscalcCalculated: 10 April 2026 to 14 October 2026, 187 days.[3, 1]

How to read this table: a grade A value is read by TrustMRR from Stripe, the payment processor, through an API key. A grade C value is a founder statement, a company page, a price list or an encyclopedia article. A "calc" value was computed on this page from the rows above it.

Koe plans in October 2026, from the company's pricing page and home page (grade C)
PlanPer month, billed yearly (US$)Per month, billed monthly (US$)Limit
Free00720p with watermark; still frames only, motion needs a plan
Hobby1012800 credits a month, 1080p with watermark
Starter24302,500 credits a month, no watermark
Pro908,000 credits a month
Team24022,000 credits a month, 4K, priority queue
Ultra60060,000 credits a month, 4K, API access
Enterprisecustomcustom LoRA character training, single sign-on
Credit packs5.99 to 109.99 one-time100 to 3,000 credits, never expire

The pricing page shows monthly prices and says that yearly billing saves 20%. The home page shows Hobby at US$10 and Starter at US$24 a month billed yearly, which saves US$24 and US$72 a year. The yearly prices of the larger plans load by script and were not read. A 6-second scene on the entry model, LTX 2.3 Pro, uses about 80 credits including its start frame; Kling and Seedance scenes cost more. Credits are spent when a generation starts and are not refunded. One-time credit packs never expire.

02

The timeline

From a Product Hunt launch with 6 upvotes to US$10,000 in MRR in three months, then a fall by three quarters. Each entry carries its grade.

  1. 28 February 2025CThe founder opens the X account @xuan_xuan_hh. Its bio says: founder, math olympiad medalist, ex-Meta. [6]
  2. 8 April 2025CThe founded date that the founder entered on TrustMRR. Stripe records no revenue for the next twelve months. [1]
  3. 18 March 2026CKoe's X account @koedotsh is created, with New York City as its location. [5]
  4. 10 April 2026CLaunch on Product Hunt as Koe, AI Film Studio: type one sentence, get a complete cinematic film. 6 upvotes, 68th of the day. [3, 4]
  5. 14 April 2026CKoe adds Seedance 2.0 and Kling Omni v3. ByteDance had started its own consumer version of Seedance 2.0 outside China on 26 March, in CapCut, and the United States was not among the first markets. [5, 21]
  6. June 2026ARevenue reaches US$7,177 in the month, up from US$645 in May. [1]
  7. 16 July 2026CThe founder posts that Koe has reached US$10,000 in MRR and that monthly revenue grew 10 times in less than a month. The post had 289 views. [2]
  8. July 2026ABest month: US$32,697 of verified revenue. ByteDance releases Seedance 2.5 on 31 July. [1, 21]
  9. 17 August 2026CKoe 2.0 is announced, available exclusively with Seedance 2.0 models. August revenue is US$20,414. [5, 1]
  10. 15 September 2026CPlatform Terms and Author Terms, version 2026-09-15. Koe offers to turn a book into a season of 18 to 24 episodes for US$499 and to sell it on a storefront at koe.show, keeping 20% of each sale. [10, 11, 14]
  11. September 2026ARevenue falls to US$8,616 in the month. [1]
  12. 24 September 2026COpenAI shuts down the Sora API, after it closed the Sora app on 26 April. [20]
  13. 14 October 2026AMRR is US$7,352 from 360 active subscriptions, and total verified revenue is US$73,304. Koe is not listed for sale. [1]
03

Why it worked

Four causes, in the order they mattered. Each one is a reusable lesson.

01

An editor on top of rented models

Koe does not train a video model. It writes the storyboard, keeps characters consistent, and sends each scene to LTX, Kling or Seedance. When a provider releases a better model, Koe adds it within days: Seedance 2.0 and Kling Omni v3 on 14 April 2026, and Koe 2.0, announced as available only with Seedance 2.0 models, on 17 August.

02

Timing with a new model

Koe offered Seedance 2.0 from 14 April 2026. ByteDance's own consumer version, started outside China on 26 March, did not include the United States in its first markets. Revenue rose from US$645 in May to US$32,697 in July. No source links the two; the dates are the only evidence.

03

Free still frames, paid motion

The free plan makes storyboards and still frames, and motion clips need a paid plan. The user sees the look of the film before paying for the expensive step, the generation of video.

04

Credits spent at the start of each generation

Plans give a monthly number of credits, and packs of 100 to 3,000 credits can be added. Credits are used when a generation starts, whether or not the user keeps the result, and they are not refunded.

How it was built

These details come from Koe's own pages: the privacy policy, the pricing page and the terms. They are grade C. [9, 7, 10, 3, 5, 6, 8]

  • The site is a Next.js web app; its page source shows the Next.js markers. It runs on Vercel and Railway.
  • Supabase stores account data and signs users in. Cloudflare R2 and the Cloudflare network store and serve the videos.
  • Stripe takes the payments. PostHog and Sentry measure usage and errors. Meta Pixel and the Conversions API measure the ads for the author program.
  • AI model providers for frames, videos, audio and assistant answers: ByteDance/BytePlus (Seedance), OpenAI, Anthropic, Google, fal.ai, Lightricks/LTX and ElevenLabs.
  • Video models on the plans: LTX 2.3 Pro as the entry model, Kling 3.0 and Kling O3, and Seedance 2.0. Enterprise adds LoRA training, a small add-on model trained on a customer's own character images.
  • Public channels: a Product Hunt launch, X posts by the founder and by the company with 18 to 388 views each, a Discord server, a subreddit and a referral program. No source says which channel brought the July customers.

Tools named in the sources

Each one opens its page in the Tools tab, with every story that uses it.

BytePlusCKlingCLightricks LTXCOpenAI APICAnthropic APICfalCElevenLabsCNext.jsBSupabaseCVercelCRailwayCCloudflare R2CCloudflareBStripeAMeta adsCProduct HuntAXCRedditCDiscordCTrustMRRAPostHogCSentryC
04

The dead that came before

Five companies in the Loot Drop database promised that anyone could make good video without skills, or tried to sell stories as short paid video. The values below are from that database and are not independently verified.

Died 2009 to 2012Burned US$500K

HighlightCam

The promise

Professional-looking video edits, made automatically by an algorithm.

Why it died

Smartphone apps and free editing software did the same job, and the market did not see the need.

CompetitionLoot Drop entry ↗
Died 2007 to 2010Burned US$20K

Fuzzwich

The promise

Simple animation for people with no animation skills.

Why it died

It stayed on the desktop while users moved to phones, and Adobe had more money.

CompetitionLoot Drop entry ↗
Died 2011 to 2014Burned US$30M

Viddy

The promise

Filters that make phone videos look professional.

Why it died

Instagram added video in 2013 and Vine spread fast; Viddy lost its users to both.

CompetitionLoot Drop entry ↗
Died 2009 to 2017Burned US$5M

Vook

The promise

Books combined with video.

Why it died

Readers kept reading plain e-books, and publishers did not adopt the format.

No Market NeedLoot Drop entry ↗
Died 2018 to 2020Burned US$1.75B

Quibi

The promise

Short premium episodes for phones, behind a paywall.

Why it died

Lockdowns removed the commute it was designed for, and free short video on TikTok kept the audience.

No Market NeedLoot Drop entry ↗

What changed

HighlightCam, Fuzzwich and Viddy promised that anyone could make good-looking video without skills. They had to build the effects themselves, and larger platforms (phone apps, Adobe, Instagram) added the same features for free. Vook and Quibi tried to sell stories as video, and readers and viewers did not pay for the format. Koe's offer to turn books into paid video seasons, in terms dated 15 September 2026, is the same bet as Vook and Quibi, with AI cutting the production price to US$499 a season.

What is different: the hard part, the video model, is now rented by the second from companies that compete to sell it, so one person can build the editor around it. The same fact is the risk. Google, Runway and Kuaishou sell the same or similar models directly, and Runway includes Kling 3.0 and Seedance 2.5 in its own plans. An editor on top of rented models keeps its customers only while its workflow is better than the model makers' own apps.

05

Can one person copy it now?

The method, yes. The July spike, no.

One person built a storyboard editor on top of rented video models and reached about US$10,000 in MRR three months after launch. The spike came with new models and faded within two months. What remains, US$7,352 a month from 360 subscriptions, is what the workflow itself holds.

Where it is still open: Narrow uses where a general video app is weak: music videos from lyrics, ads for small brands, series from books, and countries where a new model is not yet sold directly.

  1. 1Build the workflow that the model makers do not build: storyboard, consistent characters, a timeline and export.
  2. 2Sell prepaid credits priced per model and per second, and spend them when the generation starts.
  3. 3Give cheap still frames for free and charge for the expensive motion step.
  4. 4Add each new model within days of its release and post about it, but do not depend on one model.
  5. 5Connect Stripe to a public revenue page, so that readers and buyers can check the numbers.

What to watch

The risks are the mirror image of the causes above.

Revenue fell by three quarters after July

Verified revenue was US$32,697 in July 2026, US$20,414 in August and US$8,616 in September, 73.6% below July. MRR fell less, from about US$10,100 on 16 July (founder's screenshot) to US$7,352 on 14 October, minus 27.2%. No source explains the fall. Possible causes that this page could not verify: a launch spike of one-time credit purchases, users moving to other apps after Seedance 2.5 (31 July) and wider Seedance access, or cancelled subscriptions. In the last 30 days revenue averaged US$246 a day and in the last 7 days US$269 a day, so the fall has slowed to the level of the subscriptions.

The model makers sell to the same users

Google sells Veo 3.1 and its Flow video studio inside Google AI plans: AI Plus at US$4.99 a month with 200 Flow credits, AI Pro at US$19.99 with 1,000 Flow credits and a limited Veo 3.1 Lite trial, and AI Ultra from US$99.99 with Veo 3.1. Runway's Standard plan costs US$12 a month billed yearly and includes Kling 3.0, and its Pro plan lists Seedance 2.5. Kuaishou runs Kling as its own service. Runway raised US$315 million in February 2026. OpenAI closed the Sora app on 26 April 2026 and the Sora API on 24 September 2026, so even a large maker can leave this market.

Video generation is paid by the second

Veo 3.1 costs US$0.05 to US$0.60 a second on the Gemini API, and Kling v3 Pro US$0.14 a second on fal.ai; fal.ai bills Seedance 2.0 by tokens. A 6-second scene on Koe's cheapest model brings Koe US$0.80 to US$4.79, depending on the plan or pack. Koe does not publish the credit price of Kling and Seedance scenes, so its margin cannot be computed. TrustMRR shows a 93.0% profit margin for the last 30 days but does not say which costs it counts. Credits are not refunded when a user discards a result, so the user pays for failed generations.

Dependence on model providers

Koe 2.0 was announced as available exclusively with Seedance 2.0 models. In February 2026 Disney and Netflix sent ByteDance cease-and-desist letters over Seedance 2.0, and in March two US senators asked ByteDance to shut it down. If a provider raises prices, restricts access or closes a model, as OpenAI did with the Sora API, Koe must move its users to another model, and its product changes with it.

Platform rules on AI video and likeness

YouTube requires creators to disclose realistic AI-generated content, and its inauthentic content policy, renamed and clarified on 15 July 2025, excludes mass-produced or repetitive videos from monetization; Koe's pricing page tells users they can earn YouTube ad revenue. TikTok labels AI-generated content automatically when it is uploaded from some other platforms with C2PA Content Credentials, metadata that records how a file was made. Meta adds AI info labels. Koe's own Platform Terms forbid real people's likenesses or voices without their written permission.

Little evidence about the team and the new product

Only one person is named: the founder, who writes 'solo founder of Koe' on Product Hunt. The careers page lists open research and engineering roles and says the team is small, but no source gives a headcount. The book-season product is sold under terms dated 15 September 2026, and TrustMRR does not split revenue by product. The home page says Koe was featured in Rolling Stone, Billboard, Wired and Variety; this page could not verify that.

Sources

The grade next to each source is the best grade that the source can support. "Read in full" means that the text of the page was downloaded and the numbers were checked in it.

  1. A TrustMRR profile of Koe (listed as Hong Xuan Kow), revenue verified from Stripe, synced 14 October 2026
    Read the markdown profile in full and compared it with a copy cached on 12 October 2026. Revenue, MRR and subscriptions come from the Stripe API; the description, founded date and pricing text are written by the founder (grade C).
  2. C The founder's X post, 16 July 2026: Koe has reached 10k in MRR
    Read in full through api.fxtwitter.com, with the attached screenshot of the Stripe dashboard.
  3. C Product Hunt: Koe, AI Film Studio, launched 10 April 2026
    Read the page and its page data. Source for the launch date (scheduled 10 April 2026, 00:01 Pacific time) and the description.
  4. C Hunted.space launch record: Koe, AI Film Studio
    Read in full. Source for the 6 upvotes, the 68th place of the day, the name of the person who submitted it and the maker's comment.
  5. C Koe's X account @koedotsh, posts from April to August 2026
    Read the profile and its 11 visible posts through api.fxtwitter.com, two pages.
  6. C The founder's X profile @xuan_xuan_hh
    Read the profile through api.fxtwitter.com. The full timeline returned an error, so only the 10 posts with images or video were read.
  7. C Koe pricing page, read 14 October 2026
    Read in full, including the answers to its questions in the page data.
  8. C Koe home page, read 14 October 2026
    Read in full, including the answers to its questions in the page data.
  9. C Koe Privacy Policy
    Read in full. Source for the AI model providers and the other service providers.
  10. C Koe Platform Terms, version 2026-09-15
    Read in full. Source for the credit rules, the storefront split and the likeness rule.
  11. C Koe Author Terms, version 2026-09-15
    Read in full. Source for the book seasons at US$499, the legal entity Koe Platforms and Delaware law.
  12. C Koe About page
    Read in full.
  13. C Koe Careers page
    Read in full. It lists open roles but gives no headcount.
  14. C koe.show, the storefront for author-owned series
    Read in full.
  15. C Y Combinator company directory, yc-oss mirror
    Searched all 6,297 company records for Koe and koe.sh. No match.
  16. C Google Gemini API pricing
    Read the Veo 3.1 section, prices per second.
  17. C fal.ai pricing
    Read the list of video model prices.
  18. C Google AI plans
    Read the plan cards for AI Plus, AI Pro and AI Ultra.
  19. C Runway pricing
    Read the plan cards and their model lists.
  20. C Wikipedia: Sora (text-to-video model)
    Read the article text through the Wikipedia API. Source for the shutdown dates.
  21. C Wikipedia: Seedance
    Read in full through the Wikipedia API. Source for the release dates, the copyright disputes and the CapCut rollout.
  22. C Wikipedia: Veo (text-to-video model)
    Read in full through the Wikipedia API. Source for Veo in the Gemini app and in Google Flow.
  23. C Wikipedia: Kling AI
    Read in full through the Wikipedia API.
  24. C Wikipedia: Runway (company)
    Read in full through the Wikipedia API. Source for the funding rounds.
  25. C YouTube Help: Disclosing use of GenAI content
    Read in full.
  26. C YouTube Partner Program policies
    Read the section on inauthentic content and its 15 July 2025 update.
  27. C TikTok newsroom, 9 May 2024: Partnering with our industry to advance AI transparency and literacy
    Read in full. The TikTok help page on AI-generated content loads by script and could not be read.
  28. C Meta newsroom: Our Approach to Labeling AI-Generated Content and Manipulated Media
    Read in full, with its 1 July 2024 update.
  29. C Loot Drop database export, failed startups
    Local export. AI-written summaries, not independently verified. Source for the failed predecessors.

More in Video generation

Other businesses in the same group, best evidence first. All 10 stories in the group.

AI generator of faceless short and long videos · USA

Vid.AI

Evidence A
US$826,663still running, revenue falling · 25 months after launch
The product

A web application that turns an idea or a script into a finished video in which no person appears on camera: a script written by a language model, a synthetic voice, stock footage or AI images, captions and music. It makes vertical shorts and horizontal videos of up to 20 minutes, and it can upload to a YouTube channel that the customer connects. Customers pay a monthly or yearly subscription that includes credits. There is no free plan and no free trial.

Why it worked

A developer in Bangalore formed a company with an American YouTuber who already had an audience of people who want faceless channels, so customers paid from the first day. Affiliates receive 33% of every payment and brought about a tenth of all revenue, and Meta ads brought more. Nobody could use the product without paying, so no free user consumed paid AI models. Verified revenue reached about US$85,000 a month for six months, and then fell by half.

Evidence strength3/4
Copyable pattern2/4
Capital needed2/4
Launch September 2024Status October 2026Channel Founder's posts and audience Read the story
Faceless short videos made and posted on a schedule · France

StoryShort

Evidence A
US$216,036still running, revenue falling · 27 months after launch
The product

An AI tool that writes, makes and posts short "faceless" videos for TikTok, YouTube Shorts and Instagram Reels: script, AI images or clips, voiceover, music and captions. It also makes long videos of 8 to 10 minutes, AI UGC ads and talking avatars, and it can rebuild a video that did well for another creator.

Why it worked

It sells a channel instead of a single video: a series picks the topic, makes the video and posts it every day. Growth came from ads and from affiliates who keep 30% of each payment for life, and the founder swaps in new video models when better ones appear.

Evidence strength3/4
Copyable pattern2/4
Capital needed2/4
Launch July 2024Status October 2026Channel Paid ads Read the story
AI influencer and product-image generator · Singapore

The Influencer AI

Evidence A
US$50,824still running, revenue growing · 17 months after launch
The product

A web app that creates one AI person, from chosen traits or from 1 to 4 photos of the user, and then generates photos, talking videos, outfit try-ons and product shots of that same person for Instagram, TikTok, ads and online shops. Plans cost US$19 to US$199 a month and include credits that each image or video second uses up.

Why it worked

General image generators make a new face with every prompt, so a brand or creator cannot build a recognizable character with them. This app saves the face once and generates everything else from it. One developer found users through AI directories, free single-feature tools and, later, guides written for search. Credits make the model costs grow with use.

Evidence strength3/4
Copyable pattern2/4
Capital needed1/4
Launch 8 May 2025Status October 2026Channel Launch sites and deal sites Read the story

Built with the same tools

Businesses whose sources name tools that Koe also uses: OpenAI API, fal, Next.js, Cloudflare. Every tool has its own page in Tools.

Open-source social media scheduler for people and AI agents · Israel, company registered in Hong Kong

Postiz

Evidence A
US$3Mstill running, revenue growing · 27 months after launch
The product

A calendar that publishes posts to more than 30 social networks: X, LinkedIn, Instagram, TikTok, YouTube, Reddit, Bluesky, Discord and others. The whole code is open source, so anyone can run it on their own server for free; the company sells the hosted version from US$29 a month. Since 2025 AI agents such as Claude, ChatGPT and OpenClaw can connect to it through an MCP server, a command-line tool and a public API, so an agent drafts and schedules the posts and a person approves them in the calendar.

Why it worked

The founder took a twenty-year-old product category with no open-source entry and open-sourced all of it, which gave him millions of Docker downloads and 37,000 GitHub stars without a marketing budget. Revenue stayed small for a year. It multiplied when he turned the product toward people who automate their posting with n8n workflows and, from late 2025, with AI agents, which need an API-first scheduler and keep paying because they post without human effort.

Evidence strength3/4
Copyable pattern2/4
Capital needed1/4
Launch July 2024Status October 2026Channel Open source Read the story
AI designer for mobile app screens · USA

Sleek

Evidence A
US$246,516still running, revenue falling · 11 months after launch
The product

A web app that designs the screens of an iPhone or Android app from one sentence or a reference image, with AI. You chat to change the design, then paste it into Figma as editable layers, export HTML or React code, or send it to an app builder. AI coding agents can also use it through an API.

Why it worked

It took a narrower job than the general AI design tools, mobile apps only, for people who build apps without design skills. The three founders reused what they had learned from four earlier design tools, built it in three weeks, and showed finished designs on X before the launch. Each new Gemini model gave them a better result and a new demo to post.

Evidence strength3/4
Copyable pattern2/4
Capital needed1/4
Launch 17 November 2025Status October 2026Channel Founder's posts and audience Read the story
AI assistant inside messaging apps · Switzerland

Notis

Evidence A
US$161,448still running, revenue falling · 35 months after launch
The product

An AI assistant that you reach by message. You send a voice note, a text, a screenshot or a forwarded email on WhatsApp, Telegram, iMessage, Slack or email, and it writes notes, tasks, draft posts and CRM updates into Notion and more than 1,000 other connected apps. It started as a voice-to-Notion tool. In 2026 it added a desktop app, a cloud computer, automations, and connections to Claude, ChatGPT and coding agents.

Why it worked

It reached people inside the chat app that they already open all day, so capturing an idea on a phone took one voice note. It aimed first at Notion users, whom the company told that ideas get lost on the phone. Ads on Facebook and Google then paid for fast growth in early 2026, until the founder stopped them in August 2026 because the costs were larger than the revenue.

Evidence strength3/4
Copyable pattern2/4
Capital needed2/4
Launch November 2023Status October 2026Channel Paid ads Read the story

Back to all stories