CodeGuide
What is not verified
One interview, and no figure since. Every revenue figure on this page rests on one source: a Starter Story video interview with the founder, CJ Zafir, by Pat Walls, published on 21 May 2025 and recorded in about the third week of March 2025. In it the founder said that CodeGuide reached US$42,000 in MRR within 90 days of launch and projected about US$90,000 for that month; the US$17,000 a month at day 49 is the interviewer's account of a message from the founder. He called the figure both MRR and "42k per month" and did not say how yearly plans were counted. No payment processor, buyer or journalist has confirmed any of these figures. On 24 October 2026 this page checked again: TrustMRR had no CodeGuide profile (the address returned "Not found"), a search of Hacker News through the Algolia API found no story or comment about codeguide.dev, the Product Hunt page of the March 2026 relaunch gives no revenue, and a web search found only podcast directories that republish the same Starter Story episode, which is the same source and not a second one. On 18 September 2025 he posted only the words "$1M ARR.", with no product named and no screenshot; this page does not count it. A reader could look for a payment-processor profile (TrustMRR or a Stripe screenshot that names CodeGuide) or a second dated interview with a revenue figure. The headline number is grade D: one claim or one estimate, with no second source. Aiolum shows it because the story is useful, and marks it so that it is not mistaken for a verified figure.
CodeGuide is a planning tool for people who build software with AI coding tools, programs such as Cursor, Claude Code or Lovable in which a large language model writes the code from instructions in plain English. Its founder's starting point was that these tools "hallucinate" and "assume things". A hallucination is output from a model that looks right but is invented; here, code that ignores what the project needs. CodeGuide takes the user's description of the app idea and writes a blueprint for the coding tool: the features, the tech stack, the architecture, the database, the screens and a step-by-step task list, in documents such as a PRD, a product requirements document that lists what the app must do and for whom. The user attaches the documents to the coding tool as context, the text that a model reads before it answers. CodeGuide can also import an existing codebase from GitHub and plan new features for it.
CJ Zafir, who says he moved to Canada in 2018, ran an agency that built first versions of apps and spent nine to ten hours with ChatGPT on the planning documents for each one. He turned his prompts into an automated sequence and cut that to about two hours. In 2024 he set himself 12 startups in 12 months; of the 11 he built, he wrote on X in March 2025 that 3 were sold for US$30,000 in total and 7 failed. He had posted on X about AI prompting since 2022, so he posted a demo of the new idea before it existed and collected more than 1,800 waitlist emails in two weeks. With a friend who wrote the back end, he built CodeGuide in 14 days and launched it on 24 December 2024. In a Starter Story video interview published on 21 May 2025 and recorded in about March 2025, the founder told the interviewer that it reached US$42,000 in MRR, monthly recurring revenue, the amount that subscribers pay each month, within 90 days, on costs of about US$3,500 a month, most of it for the OpenAI API, the paid interface through which a program uses a provider's models.
The revenue figures on this page rest on one source, so the page is grade D, reported but not verified: the founder told Starter Story, in the interview published in May 2025, that CodeGuide reached US$42,000 a month in recurring revenue. No payment processor, buyer or journalist has confirmed it, and when this page checked on 24 October 2026, TrustMRR listed no profile for CodeGuide. His posts from the same weeks agree on the direction: he wrote on 7 March 2025 that the company was profitable on a monthly bill of US$3,495, and his user counts rose from 3,013 on 27 February to 5,019 on 23 March 2025. He has published no later revenue figure that names the product. On 24 October 2026 the site was still live, with the same plans and the same claim of more than 40,000 users as nine days before, and a relaunch on Product Hunt on 10 March 2026 shows that the product was still being developed; the founder's own posts are now mostly about fine-tuning small language models and about SkillTune, his new product. The page also shows what happened to earlier companies that sold a helpful layer on top of programmers' main tools.
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.
| Fact | Value | Grade | How we know | Sources |
|---|---|---|---|---|
| Monthly recurring revenue within 90 days of launch | US$42,000 | D | Founder in a Starter Story video interview published 21 May 2025: "in just last three months we've grown CodeGuide to 42k MRR". He also called it "42k per month" and did not say how yearly plans were counted. Day 90 after the launch was 24 March 2025. | [1] |
| Run rate from that figure | US$504,000 a year | calc | Calculated: US$42,000 times 12. The interviewer, Pat Walls, says "over 500 000 in ARR". | [1] |
| Revenue per month 49 days after launch | US$17,000 | D | The interviewer's account of a message from the founder. Day 49 was 11 February 2025 (calculated from the launch date). No money was spent on marketing. | [1] |
| Projection given in the interview | about US$90,000 for that month | D | Founder: "we're projecting to cross around 90k this month". No later figure confirms it. | [1] |
| Members at the time of the interview | more than 4,800 | C | Founder in the interview. His posts give 4,678 on 19 March and 5,019 on 23 March 2025, which places the recording at about 20 to 23 March 2025. That date is an inference on this page, not a stated date. | [1, 7] |
| Launch | 24 December 2024, 100 users on the first day | C | Founder in the interview. The product took 14 days to build. | [1] |
| Waitlist before the product existed | more than 1,800 emails in two weeks | C | Founder. The landing page was made with Bolt in about 30 minutes. His demo post got about 400 reposts and 300 sign-ups. | [1] |
| Monthly costs, February 2025 | US$3,495 | C | Founder's post of 7 March 2025: OpenAI API US$2,890 ("98% AI work"), Claude API US$340, Gemini Flash API free, and 11 other tools and services. He added: "And yes we are profitable." | [2] |
| Share of model APIs in the February 2025 costs | 92.4% | calc | Calculated: (US$2,890 + US$340) / US$3,495. The OpenAI API alone was 82.7%. | [2] |
| Monthly costs, June 2025 | US$3,477 | C | Founder's post of 17 June 2025: OpenAI API US$2,134 and Claude API US$780. He wrote that 84% of all costs were model API costs; (2,134 + 780) / 3,477 = 83.8%. | [3] |
| Costs as a share of the March 2025 revenue | about 8% | calc | Calculated: about US$3,500 a month in costs (founder, interview) divided by US$42,000 = 8.3%. | [1] |
| Users in 2025 | 3,013 (27 February), 5,019 (23 March), 7,600 (29 April), 11,356 (17 June) | C | Founder's posts. No post says how many of them pay. | [7, 3] |
| Users claimed on the website, October 2026 | more than 40,000 | C | The company's home page, read on 15 and again on 24 October 2026 with the same text. No source says how many of them pay. The CodeGuide Chrome extension shows 135 users. | [14, 22, 19] |
| Product Hunt relaunch | 10 March 2026, rank 15 of the day, 118 points | C | The Product Hunt page of CodeGuide, read 24 October 2026: featured on 10 March 2026, 108 followers. It gives no revenue and no team size. | [23] |
| Verified revenue (TrustMRR) | none found | C | The TrustMRR profile address returned "Not found" on 24 October 2026, and the local copy of 570 TrustMRR profiles has no entry for CodeGuide or its founder. | [24] |
| Price in March 2025 | US$29 a month, no trial; yearly plan US$199 with 40% off | C | Founder in the interview ("we're not doing any trials because we're using top of the line AI models") and his posts of 23 March and 29 April 2025. | [1, 7] |
| Team in March 2025 | 2 people | C | The founder did the front end and the marketing; a friend did the back end and the customer service. No later source gives the team size. | [1] |
| Outside funding | none | C | On 14 July 2025 he wrote that top venture firms wanted to invest and that he was "strongly against raising funds". | [5] |
| His other products in 2024 | 11 built: 3 sold for US$30,000 in total, 7 failed, 1 profitable (CodeGuide) | C | Founder's post of 24 March 2025. In the interview he said that 10 of the 11 collapsed. | [4, 1] |
| Followers on X, October 2026 | 64,357 | C | Profile data returned by the X API mirror on 15 October 2026. | [13] |
| Time from launch to the US$42,000 figure | 3 months | calc | Calculated: December 2024 to March 2025. | [1] |
How to read this table: a grade D value is a revenue figure that rests on one source, the founder's Starter Story interview of May 2025, and that no other source confirms. A grade C value is a statement by the founder in that interview or in a post on X, or a figure on the company's own website or its Product Hunt page. A "calc" value was computed on this page from the rows above it. No value is grade A or B.
| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Free | 0 | 0 | 120 credits once, 1 project |
| Pro | 24 | 29 | 5,000 credits a month, 15 projects; US$288 a year |
| Founder | 29 | 6,000 credits a month, 18 projects, own API keys, team members; US$348 a year |
The page shows the prices billed yearly; its monthly switch loads by script. The Pro plan's monthly price comes from the page's own button text, "pay $29", which lists the Pro plan's features; the Founder plan's monthly price was not shown. In early 2025 there was no free plan and no trial: the founder said that a month cost US$29, with 40% off the yearly plan (US$199 a year in March and April 2025), and in August 2025 he announced a new price of US$49 a month. Use is counted in credits, and Pro users can buy more credits. The plans and prices on the home page were the same on 15 and 24 October 2026; the address /pricing returns "This page could not be found".
The timeline
From an agency owner's prompts to a planning tool that claims 40,000 users, in 22 months. Each entry carries its grade.
- 2018CZafir moves to Canada and takes his first office job, which he says was a bad experience. He then builds agencies and software products. [1]
- 2022CHe writes many posts on X about prompting AI models. [1]
- 2024CHe sets out to build 12 startups in 12 months. He builds 11: 3 are sold on Acquire.com for US$30,000 in total, 7 fail, and one becomes CodeGuide. [4]
- Before the launchCHe turns the prompts that he used for his agency's planning documents into a sequence on Make.com, posts a demo, and collects more than 1,800 waitlist emails in two weeks. He and a friend build the product in 14 days. [1]
- 24 December 2024CLaunch, with 100 users on the first day. [1]
- About 11 February 2025DUS$17,000 a month after 49 days, by the interviewer's account. [1]
- 7 March 2025C3,689 members after 74 days. He posts his February bill, US$3,495, and writes that the company is profitable. [2]
- 16 March 2025CHe writes that CodeGuide uses a mixture of models, OpenAI's o1 and o3-mini and Anthropic's Claude Sonnet 3.7, to write the documents. [6]
- About 24 March 2025DUS$42,000 in monthly recurring revenue within 90 days of launch, by the founder's account in the interview. [1]
- 21 May 2025CStarter Story publishes the interview. On 16 June he writes that 100,000 people have watched it. [1, 12]
- 17 June 2025C11,356 users and a monthly bill of US$3,477, 84% of it for model APIs. [3]
- 14 July 2025CHe writes that venture firms are offering money and that he will not raise it. [5]
- 12 August 2025CCodespace, a coding agent inside CodeGuide that runs Claude Code on a server with the Kimi K2 and Claude Sonnet 4 models. He says that it does 99% of the team's own development for US$35 to US$55 a day, and announces a new price of US$49 a month. [8]
- 24 January 2026CCodeGuide 2.0: a planning and prototyping tool that writes requirements documents, wireframes, skills, prompts and tasks. [9]
- 10 March 2026CCodeGuide is featured on Product Hunt, as a tool that writes PRDs, specs and wireframes and maps an existing GitHub codebase. It ends the day at rank 15 with 118 points. [23]
- 1 August 2026CHis posts are now about fine-tuning small language models. Among the results of a 15-day break he lists "sold 1 project", without naming the project. [11, 13]
- October 2026CThe site sells a free plan and two paid plans and claims more than 40,000 users; on 24 October 2026 it is live with the same plans and prices. His X profile describes him as fine-tuning small language models and links to his new product, SkillTune. [14, 22, 13, 15]
Why it worked
Four causes, in the order they mattered. Each one is a reusable lesson.
Audience
He had posted about AI prompting on X since 2022. He showed a demo of a product that did not exist yet, collected more than 1,800 emails, and launched to them. In the interview he calls the method content-audience fit: post content every day for 14 days, read the replies to find the main problem, and send the readers to a waitlist page. All growth was organic; there was no paid marketing.
It writes the plan, not the app
CodeGuide does not write the code. It writes the documents that the coding tool needs, and the user's own Cursor, Claude Code or Lovable writes the code. That made the product small enough for two people to build in 14 days, and every new coding tool became a place where CodeGuide was useful, not a competitor.
No free trial
In March 2025 there was no free trial, because the founder said that the best models were expensive. He told Starter Story in that interview, published in May 2025, that a month cost US$29, with 40% off the yearly plan. A free plan with 120 credits came later.
Costs that stay small
The founder posted on 7 March 2025 that the February 2025 bill was US$3,495, and that US$2,890 of it was the OpenAI API. Measured against the US$42,000 a month that he told Starter Story he had reached in March 2025, the costs were about 8%. He paid affiliates through Rewardful and spent nothing on advertising.
How it was built
These details come from the founder's interview, his posts on X and the company's home page. They are grade C. [1, 2, 3, 6, 8, 14]
- Two people built the first version in 14 days: the founder did the front end and the marketing, a friend did the back end and the customer service.
- The first version was a sequence of the founder's own prompts on Make.com that cut a nine-hour planning task to about two hours. The first landing page was made with Bolt in about 30 minutes.
- Next.js for the front end; Supabase (PostgreSQL) for the database and storage; Vercel for deployment; a Hetzner server; Clerk for login; ConvertKit for email; Stripe for payments; Rewardful for affiliates.
- The code is written with AI tools: Cursor as the main editor, Windsurf, and Claude for designing interface components; by June 2025 also v0 and Bolt.
- The documents are written by a mixture of models. In February 2025 the OpenAI API did 98% of the AI work, by the founder's bill, with Anthropic's Claude doing the rest; Google's Gemini Flash API, free at the time, ran a feature called Codie. The site now offers more than 40 models, and the Founder plan lets users bring their own API keys.
- From August 2025 the team wrote almost all of its own code with Codespace, its agent built on Claude Code with Kimi K2 and Claude Sonnet 4; the founder wrote on 12 August 2025 that it cost US$35 to US$55 a day.
Tools named in the sources
Each one opens its page in the Tools tab, with every story that uses it.
The dead that came before
Five companies in the Loot Drop database sold tools that made building software easier, and failed when a bigger platform built or bought the same thing, or when users would not pay. The values below are from that database and are not independently verified.
Kite DevTool
The promiseAI code completion inside the programmer's editor.
Why it diedFew users paid for its free product, and GitHub Copilot, built on OpenAI's Codex model, gave better suggestions with GitHub's reach.
Builder.ai
The promiseCustom apps for non-technical founders, built by an AI-powered assembly line.
Why it diedIt spent US$445 million selling apps for about US$20,000 that cost about US$35,000 to build, with offshore developers doing work that was sold as AI.
CodeSee
The promiseMaps of a codebase, so that developers understand it faster.
Why it diedDevelopers liked the maps, but engineering managers saw no clear return for an occasional need, and free tools did much of the same.
Fig DevTool
The promiseEditor-style autocomplete and visual command builders for the terminal.
Why it diedFew users paid in a market full of free tools. AWS acquired it in 2024.
Pixate
The promiseMobile app prototypes without code, between design and development.
Why it diedGoogle bought it and gave it little attention, while Figma and other web-based design tools took the market.
What changed
Kite and Fig added a helpful layer to a programmer's main tool, and the owner of a bigger platform built or bought that layer: GitHub Copilot replaced Kite, and AWS bought Fig. CodeSee sold an understanding of large codebases that managers would not pay for. Builder.ai sold finished apps to non-technical founders and, by the Loot Drop record, spent more on each app than it charged. Pixate sold prototypes between design and code and lost to Figma.
CodeGuide differs in cost and in sales. It sells only the plan, not the app, so its costs were about 8% of its revenue in March 2025 by the founder's own figures as he stated them to Starter Story, and it sells to individuals through the founder's posts, with two people and no venture money. It shares the main risk of Kite and Fig: the coding tools that it feeds have started to write their own plans.
Can one person copy it now?
The method, yes. The same product, probably not.
The way CJ Zafir found and sold CodeGuide can be repeated by anyone with an audience of builders: show a demo before the product exists, collect emails, build only after people sign up, and charge from the first day. The gap that CodeGuide filled in early 2025 is closing, because the coding tools now write their own plans.
Where it is still open: Tools that prepare input for AI agents in a field where the agent lacks context and the agent's maker is not building the same thing: test plans, data pipeline specifications, a company's own rules. The founder's next product, SkillTune, writes and tests skill files, instructions that an AI agent loads for one kind of task.
- 1Do the slow task by hand for clients first, then turn your prompts into an automated sequence and measure the time saved: his went from nine hours to about two.
- 2Post a screen-recording demo before you build, and send readers to a waitlist page: he collected 1,800 emails in two weeks.
- 3Build the first version in two weeks with one partner, and charge for a subscription from launch day.
- 4Know your model cost per customer: the founder told Starter Story in May 2025 that his whole bill was about US$3,500 a month when revenue was US$42,000.
What to watch
The risks are the mirror image of the causes above.
What is not verified
One interview, and no figure since. Every revenue figure on this page rests on one source: a Starter Story video interview with the founder, CJ Zafir, by Pat Walls, published on 21 May 2025 and recorded in about the third week of March 2025. In it the founder said that CodeGuide reached US$42,000 in MRR within 90 days of launch and projected about US$90,000 for that month; the US$17,000 a month at day 49 is the interviewer's account of a message from the founder. He called the figure both MRR and "42k per month" and did not say how yearly plans were counted. No payment processor, buyer or journalist has confirmed any of these figures. On 24 October 2026 this page checked again: TrustMRR had no CodeGuide profile (the address returned "Not found"), a search of Hacker News through the Algolia API found no story or comment about codeguide.dev, the Product Hunt page of the March 2026 relaunch gives no revenue, and a web search found only podcast directories that republish the same Starter Story episode, which is the same source and not a second one. On 18 September 2025 he posted only the words "$1M ARR.", with no product named and no screenshot; this page does not count it. A reader could look for a payment-processor profile (TrustMRR or a Stripe screenshot that names CodeGuide) or a second dated interview with a revenue figure.
The coding tools now plan for themselves
Claude Code's changelog lists "Improvements to plan mode" in version 1.0.33, published on 23 June 2025. Kiro, an AI coding editor announced on 14 July 2025, is built around specs, documents in which a feature is thought through before the code is written. Cursor introduced Plan Mode on 7 October 2025 and wrote that most new features at Cursor now begin with its agent writing a plan. CodeGuide's answer was its own coding agent in August 2025 and wireframes, skills and prompts in version 2.0 in January 2026.
A product built on one trend
CodeGuide grew with vibe coding, which means building software by describing it to an AI tool and accepting the code that it writes, and which spread in early 2025. The founder's own record shows how short such products can live: of 11 products he built in 2024, 7 failed and 3 were sold for US$30,000 in total. Of the 255 posts of his own that the X API returned for February to mid-October 2026, one mentions CodeGuide; the rest are mostly about fine-tuning small models and about SkillTune. On 1 August 2026 he wrote that he had sold a project, without saying which. Whether CodeGuide still has the same owner and his full time is not known, and no source after March 2025 says whether the team is still two people. The product itself was still being developed in 2026: version 2.0 in January and a Product Hunt relaunch on 10 March.
Dependence on model providers
In February 2025 the OpenAI API was 82.7% of all costs, and in June 2025 model APIs were 84%. The quality and the cost of every document depend on models whose prices OpenAI, Anthropic and Google set, and Anthropic also sells Claude Code, a coding agent with its own plan mode. The founder has moved between providers several times, from mostly OpenAI to a mix with Claude, Gemini and Kimi K2.
No named company, changing prices
The website names no operating company: its terms and privacy paths return "not found", the extension's privacy policy names no company, and the Chrome Web Store lists the developer as a non-trader. Prices have changed several times: US$29 a month in early 2025, a new price of US$49 a month announced in August 2025, and US$24 to US$29 a month billed yearly in October 2026, now with a free plan, unchanged when the page was read again on 24 October 2026.
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.
- C Starter Story, 21 May 2025: From Zero to $42K/Month in 90 Days with AI (video interview with CJ Zafir by Pat Walls)Read in full: the page's automatic transcript of the video. It misspells some product names ("Cloud" for Claude, "Winserve" for Windsurf, "Superbase" for Supabase). Source for the MRR figure, the launch date, the team, the costs, the prices and the method.
- C CJ Zafir on X, 7 March 2025: monthly bill for February, 3,689 members in 74 daysRead in full through api.fxtwitter.com. Source for the February costs, the tools and "And yes we are profitable."
- C CJ Zafir on X, 17 June 2025: monthly bill, 11,356 usersRead in full through api.fxtwitter.com. Source for the June costs and the 84% share of model APIs.
- C CJ Zafir on X, 24 March 2025: 12 startups in 12 monthsRead the post in full through api.fxtwitter.com. Source for the 11 products, the 3 sales for US$30,000 and the 7 failures.
- C CJ Zafir on X, 14 July 2025: on venture firms that want to invest in CodeGuideRead in full through api.fxtwitter.com. Source for the absence of outside funding.
- C CJ Zafir on X, 16 March 2025: the models that write the documentsRead in full through api.fxtwitter.com. Source for o1, o3-mini and Sonnet 3.7.
- C CJ Zafir on X, 23 March 2025: Home of 5019 AI Coders, yearly membership US$199Read in full through api.fxtwitter.com, with the same post on 27 February 2025 (3,013, status 1895180690440012103), 19 March 2025 (4,678, status 1902395744344273163) and 29 April 2025 (7,600, status 1917237979854336027).
- C CJ Zafir on X, 12 August 2025: Codespace and the new priceRead in full through api.fxtwitter.com, with the launch post of 16 August 2025 (status 1956720628461134272) that names Kimi K2 and Claude Sonnet 4.
- C CJ Zafir on X, 24 January 2026: Introducing CodeGuide 2.0Read in full through api.fxtwitter.com.
- C CJ Zafir on X, 18 September 2025: $1M ARR.Read in full through api.fxtwitter.com. The post has no other text, no image, no quoted post and no reply target.
- C CJ Zafir on X, 1 August 2026: me 2.0Read in full through api.fxtwitter.com. Source for "sold 1 project".
- C CJ Zafir on X, 16 June 2025: 100,000 people have watched the Starter Story interviewRead in full through api.fxtwitter.com.
- C CJ Zafir's X profile and timeline, read 15 October 2026Read all 607 posts that api.fxtwitter.com returned, 26 February 2025 to 13 October 2026; the API returned no older posts. Source for the follower count, the profile text and the count of posts that mention CodeGuide.
- C CodeGuide home page and pricing, read 15 October 2026Read in full. Source for the plans, the prices billed yearly, the "pay $29" button, the features and the 40,000 users claim.
- C SkillTune home page, read 15 October 2026Read in full. The open letter on the page is signed by CJ Zafir as CEO and founder.
- C Cursor blog, 7 October 2025: Introducing Plan ModeRead in full. The company's own announcement.
- C Claude Code changelog, with the npm registry record for its versionsRead the plan mode entries. The publish date of version 1.0.33 (23 June 2025) is from https://registry.npmjs.org/@anthropic-ai/claude-code.
- C Kiro blog, 14 July 2025: Introducing KiroRead in full. The maker's own announcement of its spec-based coding editor.
- C Chrome Web Store: CodeGuide extension, read 15 October 2026Read in full. Source for the 135 users and the non-trader status.
- C CodeGuide extension privacy policy, read 15 October 2026Read in full. It names no company. The site's /terms and /privacy paths returned 404.
- C Loot Drop database export, 1,749 failed startupsLocal export. AI-written summaries, not independently verified. Source for the failed predecessors.
- C CodeGuide home page and pricing, read again 24 October 2026Read in full. The plans, prices and the 40,000 users claim are the same as on 15 October 2026. The site has no robots.txt (the address returns the site's "not found" page), and /pricing returns "This page could not be found".
- C Product Hunt: CodeGuide, read 24 October 2026Read in full. Source for the relaunch featured on 10 March 2026, the day rank, the points and the followers. An older Product Hunt page, /products/codeguide, has 1 review and 1 follower.
- C TrustMRR: CodeGuide profile address, checked 24 October 2026Could not open: the address returned "Not found". The local copy of 570 TrustMRR profiles has no entry for CodeGuide or CJ Zafir. Source for the absence of a verified revenue profile.