{
  "data": {
    "author": "Quynh Tran-Thanh",
    "faq": null,
    "kicker": "Blog · From the data · 20 November 2026",
    "lede": "Half the jobs agents ask for most have no product that finishes them, and the companies with no investor on record finish jobs about as often as the funded ones. I tested what bootstrap builders argue about against 768 agent jobs.",
    "readingMinutes": 11,
    "words": 2435
  },
  "kind": "anchor.page",
  "links": {
    "api": "https://www.anchorterminal.com/api/v1/index.json",
    "html": "https://www.anchorterminal.com/blog/the-jobs-nobody-finishes",
    "json": "https://www.anchorterminal.com/blog/the-jobs-nobody-finishes.json",
    "llms": "https://www.anchorterminal.com/llms.txt",
    "markdown": "https://www.anchorterminal.com/blog/the-jobs-nobody-finishes.md",
    "mcp": "https://www.anchorterminal.com/mcp",
    "slim": "https://www.anchorterminal.com/blog/the-jobs-nobody-finishes.min.md"
  },
  "markdown": "By Quynh Tran-Thanh, 20 November 2026. 2,435 words, about 11 min to read.\n\nIf you build alone or with a small team and no investors, 2026 has given you a lot of reasons to worry. Code is cheap, so whatever gets traction gets cloned. Every lab launch reads like the end of another category. And the advice on Hacker News is split between \"find a niche\" and \"the only moat left is money\".\n\nSo I wanted to put numbers on it. What could a self-funded builder start today that would still be standing in two years? I took our niches data, wrote down what I expected before looking, and tested it.\n\n## The research\n\nI started with what people argue about. The biggest thread in 2026 on the question was \"Domain expertise has always been the real moat\", with 884 points and 549 comments when I read it [[1]](#ref-1). The replies under it and its neighbours go round the same few positions.\n\n- Anything that works gets copied. \"You know that if anything you build gets traction, it'll be cloned by 100 people, right?\" [[2]](#ref-2)\n- The gap is the moat. \"it's having a good idea that solves a problem that nobody else is solving well\" [[3]](#ref-3)\n- Selling is the moat. \"it's not domain expertise; it's sales\" [[4]](#ref-4)\n- The labs will take it. \"Why should I believe that Anthropic will care about this product in 2, 3 years?\" [[5]](#ref-5)\n\nNobody in those threads answers with data. One commenter asks exactly that. \"How do you know \"domain expertise\" is a \"moat\"? Vibes?\" [[4]](#ref-4)\n\nThe data I used is our map of what agents try to do with the tools we list [[6]](#ref-6). Its unit is a job, so it's worth saying what that means here.\n\nA job is one specific thing a person wants an agent to get done with a tool, with a test for when it's done. It is smaller than a product or a market. Weather data is a category, and a weather API is a product. Getting the forecast for a place over the next hours or days is a job. People ask for it as \"will it rain tomorrow afternoon\" or \"give me the 10-day forecast for Lisbon\". It counts as done when one call returns hourly or daily values (temperature, rain, wind) for the whole window, with units and the place's time zone, and says when the forecast was issued [[10]](#ref-10). A product that returns the forecast in UTC, or sells the 10-day forecast separately from the hourly one, does the job with a limit.\n\nTo build the map, research agents read first-hand accounts (Hacker News comments, GitHub issues, forum posts) for each of 100 categories and wrote down the five to twelve jobs people bring to that category [[6]](#ref-6). Then each listing in the category was read against every job, from its public docs, and marked as doing the whole job, doing it with a limit, or not doing it.\n\n- 768: jobs agents bring to a category, each with a test for done\n- 100: categories\n- 3,906: first-hand accounts behind the jobs\n- 1,644: listings read against their category's jobs\n- 13,399: marks, yes, with a limit, no or unchecked\n\nThe niches data, 20 November 2026. Source [6].\n\nTwo numbers come out of it for every job. Its weight is its share of the category's work, from how often the accounts asked for it, and a category's weights add up to 100. The forecast job's weight is 30, because 24 of the 43 accounts read for the weather category asked for it [[10]](#ref-10). Its competition is how many companies' products finish it fully, counting a company once, and it reads none, one, a few (two or three) or crowded (four or more) [[6]](#ref-6). I joined that to who is behind each listing, from the ownership and investor records we keep [[7]](#ref-7), and to the year of acquisitions and shutdowns on our Sunsets page [[8]](#ref-8).\n\nBefore running anything I wrote down nine hypotheses, most of them the forum's conventional wisdom, and what result would prove each one wrong [[9]](#ref-9). Four failed outright.\n\n| What I expected | Result |\n| --- | --- |\n| Jobs nobody finishes have less demand than crowded ones | Failed |\n| Funded companies finish more jobs than unfunded ones | Failed |\n| Big companies hold the crowded jobs | Failed |\n| What stops the open jobs is access and permission (sign-in, approvals, terms) | Failed |\n| Usage-priced products finish more jobs | Mixed, the gap disappears inside a category |\n| Price is a blocker in at least a fifth of jobs | Held |\n| MCP servers finish fewer jobs than the APIs they sit on | Held |\n| Agent payments are more open than the average | Held only on paper |\n| The labs buy outside business and domain categories | Borderline |\n\n## What the data says\n\n### The empty spaces aren't empty for lack of demand\n\nThis was the one I was surest about. If nobody finishes a job, the usual reason is that nobody wants it badly enough. So I expected the crowded jobs to carry clearly more weight.\n\nThey don't. The jobs with no full fit have a median weight of 12 and five first-hand accounts behind them [[6]](#ref-6). The crowded ones have 11 and six [[6]](#ref-6).\n\nMedian demand for a job, by how many companies finish it\n\n| Companies that finish it | Median weight | Median accounts |\n| --- | --- | --- |\n| None (350 jobs) | 12 | 5 |\n| One (131 jobs) | 11 | 5 |\n| Two or three (152 jobs) | 11 | 5 |\n| Four or more (135 jobs) | 11 | 6 |\n\nEach series has a zero-based scale.\n\nIt holds where demand is highest too. In 51 of the 100 categories, the job people ask for most has no product that finishes it. Of the 139 jobs that are at least a fifth of their category's work, 75 have no full fit [[6]](#ref-6).\n\nThe forecast job from the start is my favourite example, because it sounds solved. Its 24 accounts are the most behind any open job [[10]](#ref-10). Sixteen companies do it with a limit and none does it fully [[10]](#ref-10). The blockers are free tiers withdrawn, sign-ups that need a card or a person, and forecast lengths sold as separate entitlements [[10]](#ref-10). Finding a free slot across calendars, answering a metric question from tracked events and keeping a bank connection's consent alive look the same, with 13 to 18 accounts each and no full fit [[6]](#ref-6).\n\n### Open mostly means nearly done\n\nSo why is nobody finishing them? 282 of the 350 open jobs (81 per cent) have three or more companies that do the job with a limit [[6]](#ref-6). These aren't markets nobody has entered. They're jobs several products get most of the way through.\n\nI expected the last bit to be gatekeeping, the OAuth screens and approval queues and terms that bar agents, because the accounts complain about those the loudest. I sorted the 1,091 blockers on open jobs by keyword. 63 per cent are the product not doing the job reliably, and a hand check of samples says the real share is higher, because the access group picked up things that weren't access problems [[9]](#ref-9).\n\nThey're plain things. A serverless endpoint with a cold start of a minute or more. Times returned without an offset. A row limit that caps an aggregate without saying so. A job that runs for minutes and then fails with an internal error [[6]](#ref-6). None of them needs a breakthrough. They need someone who cares about that one job.\n\n### Money doesn't predict who finishes the job\n\nThe \"only moat left is money\" camp would expect funded companies to finish more. Of the marks given to listings from VC-backed companies, 13 per cent are full fits. For companies with no investor on record it's 11 per cent, and for public companies and their subsidiaries also 11 per cent [[6]](#ref-6) [[7]](#ref-7).\n\nMarks that are a full fit, by who is behind the listing, per cent\n\n| Behind the listing | Full fit |\n| --- | --- |\n| VC-backed (705 listings) | 13 |\n| Public or a subsidiary (341) | 11 |\n| No investor on record (282) | 11 |\n| Acquired (137) | 9 |\n| Open-source project (88) | 7 |\n\nEach series has a zero-based scale.\n\nThat small gap disappears once you compare like with like. Inside the same category, listings with no investor on record do better in 27 of 53 categories and VC-backed ones in 24 [[9]](#ref-9).\n\nAnd where only one company finishes a job, it's more often a company with no investor on record than its share of listings would suggest. They are 17 per cent of listings and 24 per cent of the jobs with exactly one full fit (31 of 131) [[6]](#ref-6) [[7]](#ref-7). Ride with GPS on summarising a person's training, Late on publishing one post to several networks, Grist on appending rows to a spreadsheet and Centrifugo on pushing live updates to browsers are four of them [[6]](#ref-6).\n\nBig companies don't own the crowded jobs either. Of the companies that fully finish a crowded job, 50 per cent are VC-backed, 20 per cent public and 16 per cent have no investor on record [[6]](#ref-6) [[7]](#ref-7).\n\nMy reading, which the data shows as a pattern and not a cause, is that a full fit rewards a narrow job done completely. A focused small product can do that. A platform with forty jobs does most of them with a caveat.\n\n### Wrappers finish less than the thing they wrap\n\nThere's a running argument about whether an MCP server is a business. One commenter in 2025 was \"baffled no software vendor has already come up with a subscription to access the API via MCP\" [[11]](#ref-11). On finishing jobs, MCP servers do worse than APIs. 9 per cent of their marks are full fits against 13 per cent for HTTP APIs, and inside a category the API comes out ahead in 33 of 51 [[9]](#ref-9). An MCP server inherits the gaps of the API under it and adds its own, schemas that reject valid calls and hosts that won't start it [[6]](#ref-6).\n\n### The labs bought tools and closed products\n\nOn the lab worry, I looked at the year of exits on our Sunsets page. Of the 35 that involved a frontier lab, 29 were labs buying a company or hiring its people, and most of those were developer and agent infrastructure (Bun, Stainless, Astral, Promptfoo). The other six were labs closing products of their own, among them the OpenAI Assistants API, OpenAI Agent Builder and Project Mariner [[8]](#ref-8). Ten of the 29 deals were in a business or domain category, four of them in health and drug development, so that corner isn't safe [[8]](#ref-8). I went through the labs' shopping in more detail in [Who gets paid in the agent stack](/blog/who-gets-paid-in-the-agent-stack).\n\n## Key takeaways\n\n- Half the categories' most-asked-for agent job has no product that finishes it, and those open jobs carry as much demand as the crowded ones.\n- Most open jobs are nearly done by several products. What's missing is reliability on one job, not a new idea.\n- Companies with no investor on record finish jobs as often as funded ones, and are over-represented where only one company finishes a job.\n- An MCP server on its own finishes fewer jobs than an API.\n- The labs spent the year buying developer infrastructure and closing some of their own agent products.\n\n## What it means if you're building on your own money\n\nFor me the most useful part is how small the targets are. A job here is something like \"get tomorrow's forecast for this place\" or \"find a time that's free for everyone\", with a written test for done. That's a size one person can own completely.\n\n- **Pick a job, not a category.** Look for a job with real demand where several products get close and none finishes. The niches page lists them with their blockers [[12]](#ref-12), and each job's page names who falls short and why [[10]](#ref-10).\n- **Compete on the last 20 per cent.** The blockers are a spec. Cold starts, time zones, silent caps and failures with no error are the work. That kind of work is boring enough that funded teams with forty jobs leave it.\n- **Sell the job, not the wrapper.** If the thing you'd sell is an MCP server on someone else's API, you inherit their gaps.\n- **Be careful with selling to agents.** Paying per call and buying for a person are wide open, every job in our pay-per-call category and five of six in agent checkout [[6]](#ref-6). But they rest on three to eight accounts each, and the one first-hand report I found on charging agents through x402 says \"I implemented x402 across a very heavily crawled site and have yet to see any transactions go through\" [[13]](#ref-13). I read that as early, not neglected.\n\nRob Walling put the risk side of this well. \"If your entire product is one feature, I think you're exposed.\" [[14]](#ref-14) I agree, and the data adds that owning one job end to end is different from being one feature, because the job is where everyone else stops short.\n\nWhat the data can't tell you is whether any of these jobs pays. It counts demand in accounts, not money, and it reads products from their docs rather than testing them. It is also a draft that a person hasn't reviewed yet, it only covers companies we list, and \"no investor on record\" means we didn't find one, which isn't proof of bootstrapping [[6]](#ref-6) [[7]](#ref-7). The next thing I want to know is which of the open jobs someone is already paying to have done by hand.\n\n## Explore the niches yourself\n\nEverything in this post comes from the [niches page](/niches/), and you can use it to look for a job of your own. It needs no account or sign-in [[12]](#ref-12).\n\n1. Start with the table \"Open jobs with demand\". It lists the jobs where no company's docs show a full fit and the job is at least 15 per cent of its category's work, heaviest first [[12]](#ref-12).\n2. Or go to \"Every job, by category\" and open a category you know well. Knowing a domain is the filter this data doesn't have.\n3. Click a job. Its page shows what counts as done, how people ask for it, every company that does it with a limit and what that limit is, and what stops agents finishing [[10]](#ref-10). That last list is the closest thing to a spec for what to build.\n4. Read the reasons before calling it a gap. A job with no full fit can be open ground, or a job whose test the docs can't settle, and a product we don't list may already do it.\n5. Look at \"Jobs one company does\" to see where a single company has a job to itself. Several of them are small companies with no investor on record.\n\nIf you'd rather query it, or hand it to your own agent, the same data is an API with no key [[12]](#ref-12). This lists the jobs no company fully does that are at least a fifth of their category's work.\n\n```\ncurl 'https://www.anchorterminal.com/api/v1/niches?competition=none\u0026minWeight=20'\n```\n\nAdd `area=payments` (or any other area) to narrow it, or use `sort=demand` to rank every job by how often it's asked for. Each job is also at `/api/v1/use-cases/\u003cid\u003e.json` with every listing's mark, reason and source, and over MCP the tools are `find_niches` and `get_use_case` [[12]](#ref-12).\n\n---\n\nMore in [From the data](https://www.anchorterminal.com/blog/from-the-data/index.md)\n\n- Older: [What agents take from an agent-first site, and what they send back](https://www.anchorterminal.com/blog/what-agents-send-back.md)\n- All posts: https://www.anchorterminal.com/blog/index.md\n\nAnchor Terminal grades the tools AI agents run on, with a source for every score. [The top tools](https://www.anchorterminal.com/top/index.md) · [Categories](https://www.anchorterminal.com/categories/index.md) · [Compare two tools](https://www.anchorterminal.com/compare/index.md) · [Prices](https://www.anchorterminal.com/prices/index.md) · [Sunsets](https://www.anchorterminal.com/sunsets/index.md) · [API, MCP server and CLI](https://www.anchorterminal.com/agents/index.md)\n\n## References\n\nSources read on 20 November 2026. Points and comment counts are as Hacker News showed them that day.\n\n[1] Hacker News, \"Domain expertise has always been the real moat\", 30 May 2026. Source for the thread's points and comments. https://news.ycombinator.com/item?id=48340411\n\n[2] Hacker News, comment by AstroBen on \"The only moat left is money?\", 18 February 2026. Source for the quote on cloning. https://news.ycombinator.com/item?id=47062899\n\n[3] Hacker News, comment by showerst on \"The only moat left is money?\", 18 February 2026. Source for the quote on solving a problem nobody else solves well. https://news.ycombinator.com/item?id=47062806\n\n[4] Hacker News, comment by avaer on \"Domain expertise has always been the real moat\", 30 May 2026. Source for the quotes on sales as the moat and on vibes. https://news.ycombinator.com/item?id=48341207\n\n[5] Hacker News, comment by taylorlapeyre on \"Claude Design\", 17 April 2026. Source for the quote on whether Anthropic will care in two or three years. https://news.ycombinator.com/item?id=47807982\n\n[6] Anchor Terminal, niches data (niches.json), generated 20 November 2026. Source for the 768 jobs, their weights, accounts, competition, blockers and full fits, the 1,644 readings and 13,399 marks, and the full-fit rates by listing. https://www.anchorterminal.com/api/v1/niches.json\n\n[7] Anchor Terminal, ownership and investor records for each listing, read 20 November 2026. Source for which listings are VC-backed, public, acquired, open-source projects or have no investor on record. https://www.anchorterminal.com/investors/\n\n[8] Anchor Terminal, Sunsets, Exits, November 2025 to November 2026. Source for the 35 exits involving a frontier lab, the deals, the buyers and the six products the labs closed. https://www.anchorterminal.com/sunsets/\n\n[9] Anchor Terminal, the hypotheses, analysis script and results for this post, 20 November 2026, not published. Source for the hypotheses and their order, the comparisons inside categories, the keyword sort of blockers and its hand check.\n\n[10] Anchor Terminal, the weather forecast job, read 20 November 2026. Source for the forecast job's asks, its test for done, its weight of 30 from 24 of 43 accounts, the 16 companies with a limit, no full fit and its blockers. https://www.anchorterminal.com/api/v1/use-cases/weather.forecast.json\n\n[11] Hacker News, comment by iLoveOncall on \"MCP: An (Accidentally) Universal Plugin System\", 28 June 2025. Source for the quote on subscriptions to an API through MCP. https://news.ycombinator.com/item?id=44405296\n\n[12] Anchor Terminal, Niches, jobs few AI tool companies do well, read 20 November 2026. Source for the list of jobs and their blockers, what the page's tables show, and the API's parameters and MCP tools. https://www.anchorterminal.com/niches/\n\n[13] Hacker News, comment by ddxv on \"Brave Search API now accepts agentic payments via the x402 protocol\", 13 November 2026. Source for the quote on no x402 transactions. https://news.ycombinator.com/item?id=50479355\n\n[14] Startups For the Rest of Us, episode 842, What is the future of SaaS in an AI world, 21 July 2026. Source for Rob Walling's quote on one-feature products. https://www.startupsfortherestofus.com/episodes/episode-842-what-is-the-future-of-saas-in-an-ai-world-rob-solo\n",
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "attributionRequired": true,
    "credit": "CC BY 4.0, credit Anchor Terminal and link the page or API address you took it from (https://www.anchorterminal.com/terms/#data)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-11-20",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "1.0",
    "openapi": "https://www.anchorterminal.com/openapi.json",
    "preview": false,
    "run": "2026-10-01",
    "runLabel": "October 2026 research run",
    "terms": "https://www.anchorterminal.com/terms/#data"
  },
  "page": {
    "breadcrumbs": [
      {
        "name": "Home",
        "url": "https://www.anchorterminal.com/"
      },
      {
        "name": "Blog",
        "url": "https://www.anchorterminal.com/blog/"
      },
      {
        "name": "From the data",
        "url": "https://www.anchorterminal.com/blog/from-the-data/"
      },
      {
        "name": "The jobs nobody finishes",
        "url": ""
      }
    ],
    "description": "Quynh Tran-Thanh tests what bootstrap builders argue about (is the gap empty for a reason, does money win, will the labs ship it) against 768 jobs agents bring to 100 categories, who finishes each one, and what stops the rest.",
    "facts": [
      "Quynh Tran-Thanh",
      "20 November 2026"
    ],
    "h1": "What to build that lasts. The jobs agents ask for that nobody finishes",
    "image": "https://www.anchorterminal.com/assets/og/blog-the-jobs-nobody-finishes.png",
    "path": "/blog/the-jobs-nobody-finishes",
    "published": "2026-11-20",
    "section": "blog",
    "title": "What business to build in 2026 that survives, from 768 agent jobs | Anchor Terminal",
    "toc": null,
    "updated": "2026-11-20",
    "url": "https://www.anchorterminal.com/blog/the-jobs-nobody-finishes"
  },
  "tokens": {
    "markdown": 4850,
    "slim": 2680
  },
  "version": 1
}
