WEIR

Startup Ideas on Twitter: 6 Tested Searches (2026)

Find startup ideas on Twitter by searching what people wish existed: six tested X searches, what 92 posts really asked for, and how to spot real demand.


Published · Updated · 7 min read · 阅读中文版

Cover image: What 92 posts wished for

To find startup ideas on Twitter, search for what people wish existed, then count how many different people want the same thing.

It's for founders who want app and startup ideas with proof. On 16 November 2026, we tested six searches on X (Twitter) and read every post. We run Weir, the API behind the live steps.

Paste this into X's search box and pick Latest:

-filter:nativeretweets ("wish there was an app" OR "wish there was a tool" OR "wish there was a website" OR "wish there was an extension") -giveaway -hiring lang:en -filter:replies
  1. Run a wish searchOn Latest, in English.
  2. Mark each postA real wish, or noise.
  3. Count different peopleRepeats beat likes.
  4. Check what existsThen test a price.

6 searches that find startup ideas on Twitter

We tested each on 16 November 2026, reading up to 15 of the newest posts from 9 to 15 November (UTC). Each link opens our builder filled in, where Run it here shows 20 posts after a free sign-in.

1. "I wish there was an app"

The search at the top of this page: 8 of 15 posts were real wishes, spread over four days. Open it filled in.

Recipe card for the "I wish there was an app" search on X, with the same search as a Weir request
Recipe 1, tested 16 November 2026.

2. "Why is there no way to"

The best search for problems to solve: all 13 posts were real complaints, mostly about missing settings.

-filter:nativeretweets ("why is there no app" OR "why isn't there an app" OR "why is there no tool" OR "why isn't there a tool" OR "why is there no way to") -giveaway -hiring lang:en -filter:replies

Open it filled in

3. "So frustrating" plus a category

Swap the words in brackets for your category. Ours found 13 real complaints in 15 posts.

-filter:nativeretweets "so frustrating" (app OR software OR website) -giveaway -hiring lang:en -filter:replies

Open it filled in

4. "Someone should build"

It leans toward AI tools: 7 real ideas in 15 posts.

-filter:nativeretweets ("someone should build" OR "someone needs to build" OR "somebody should build" OR "someone should make an app") -giveaway -hiring lang:en -filter:replies

Open it filled in

5. "Is there an app that", filtered

Unfiltered, 13 of 15 posts were noise, mostly marketers discussing this tactic. The last five exclusions cut it to 7 of 11.

-filter:nativeretweets ("is there an app that" OR "is there a tool that" OR "is there an app for" OR "is there a tool for") -giveaway -hiring -reddit -lead -leads -founders -builders lang:en -filter:replies

Open it filled in

6. "I'd pay for an app"

Rare: in 30 days the first page held 3 posts, 2 of them real. The wider "I'd pay good money for" found mostly merch.

-filter:nativeretweets ("I'd pay for an app" OR "would pay for an app" OR "I'd pay for a tool" OR "would pay for a tool" OR "I'd pay for software") -giveaway -hiring -reddit -lead -leads -founders -builders lang:en -filter:replies

Open it filled in

More by goal: X search recipes.

What 92 posts asked for, 9 to 15 November 2026

We sorted every post into real wishes or noise, then grouped the real ones by theme.

92posts from 87 accounts
47real wishes or complaints
45noise, such as jokes, plugs and merch
Table of 9 themes from 47 real X posts, 9 to 15 November 2026, with posts, people and phrases per theme
Weir X search, 9 to 15 November 2026 (UTC).

We scored themes by people, not likes: one point per different person wanting the same fix, a bonus when a second phrase found it, and likes only to break ties.

Likes would have misled us. The three most-liked posts were a political jab, a webcomic plugging itself and a joke.

Most themes were loose groups of different wishes. The tightest was help finding someone to do an activity with: four people, three phrases, one week.

The biggest, 12 people, wanted settings from 10 different apps, and each company can ship its own fix.

Real wishes or complaints per search, 16 November 2026
why is there no ...13 of 13
so frustrating + app13 of 15
wish there was an app8 of 15
someone should build7 of 15
is there an app that, filtered4 of 11
is there an app that, no filter2 of 15
I'd pay good money for1 of 15

How to cut the noise on X

Each rule comes from a post we threw out:

  • Drop reposts and replies with -filter:nativeretweets -filter:replies.
  • Exclude plurals too. -giveaway missed a post that used the plural, so add -giveaways.
  • Filter the marketers. Posts about finding ideas kept using Reddit, lead and builders.
  • Skip makers' pitches. A post that prices the poster's own idea is a test, not demand.

Is it real demand or just venting?

A complaint is a clue, not a customer. Most complaints we read named one company's app, which that company can fix in an update.

Before you build:

  • Count people over weeks. Our activity-partner need had 4 people in a week; watch it for a month.
  • Search for what exists. Meetup calls itself "The people platform. Where interests become friendships."
  • Read why people reject existing apps. Both people who wanted to log what they read or watch knew apps for it.
  • Test a price. Pay talk was rare in recipe 6, so count clicks on a priced landing page.

Paul Graham wrote in 2012 that "the way to get startup ideas is not to try to think of startup ideas." Then talk to people who have the problem.

To see what launched this week, read finding new products on X. When a need checks out, find beta testers on X. Raising later? See finding investors on Twitter.

Get the top wishes every week, automatically

Ask your AI app on Mondays

With Weir connected to Claude or ChatGPT, paste this:

Find this week's most repeated wishes about <category> on X, using "wish
there was an app" and "why is there no app". Skip reposts, jokes, ads and
politics. Count different people per need and list the top 3. No names
or quotes.

For fitness, the honest answer was: no repeats yet.

AI chat answer listing this week's fitness wishes on X: 7 posts, 2 real wishes from 2 people
One Weir search, 16 November 2026.

Or run a 14-line Python script

It counts different people per theme over 7 days. Set WEIR_API_KEY and edit THEMES.

import json, os, re, urllib.request
from datetime import datetime, timedelta, timezone
THEMES = {"find a partner": {"partner", "buddy", "roommate"}, "money": {"bank", "budget", "subscription"},
          "fitness": {"gym", "workout", "walk"}, "media": {"books", "movies", "music"}}
since = (datetime.now(timezone.utc) - timedelta(days=7)).isoformat(timespec="seconds")
body = {"anyWords": ["wish there was an app", "why is there no app", "is there an app that"],
        "query": "-filter:nativeretweets", "language": "en", "onlyOriginal": True, "since": since, "limit": 20}
headers = {"Authorization": "Bearer " + os.environ["WEIR_API_KEY"], "Content-Type": "application/json"}
req = urllib.request.Request("https://api.weirapp.io/v1/x/search", json.dumps(body).encode(), headers)
posts = json.load(urllib.request.urlopen(req, timeout=60))["data"]
print(len(posts), "posts from", len({p["author"]["id"] for p in posts}), "people since", since[:10])
for name, words in THEMES.items():
    people = {p["author"]["id"] for p in posts if words & set(re.findall(r"[a-z]+", p["text"].lower()))}
    print(f"{name:15} people: {len(people)}")

Our run on 16 November 2026 printed:

20 posts from 19 people since 2026-11-09
find a partner  people: 0
money           people: 0
fitness         people: 2
media           people: 1

It matches whole words, so add plurals such as roommates to THEMES. For more pages, follow meta.nextCursor.

No code? The free Tweet Finder runs a word and date search and saves a CSV.

What does it cost?

A 20-post search costs 21 credits; 5,000 credits a month are free, and $5 buys 25,000. Our research cost 140 credits.

Free accounts can get a 429 when shared capacity is busy. It is never charged, and our longest wait was 10 minutes.

When not to use Weir

  • You'll look once a month. X's search box is free.
  • You'd rather buy a list. Paid idea databases do the reading; ask how fresh their data is.
  • Your users aren't on X. Read their forums and app reviews instead.

Weir is read-only, reads public posts only, and isn't licensed by X, so access can change. Keep notes to needs and counts, never people.

With a product, the same searches find buyers: see Twitter lead generation.

FAQ

How do I find startup ideas on Twitter?

Search X for wish and complaint phrases, like "wish there was an app", on Latest. Group the real posts by need and count the different people behind each.

What should I search on Twitter for business ideas?

Try "why is there no way to", "so frustrating" plus your category, and "someone should build". Add -filter:nativeretweets -filter:replies lang:en to each.

Is a complaint on Twitter proof of demand?

No. Most complaints we read were about one company's app. Look for the same need from different people over weeks, check what exists, then test a price.

Can ChatGPT or Claude find startup ideas from tweets?

Yes, with a tool that reads X. With Weir connected over MCP, paste the Monday prompt above.

Sources

All checked on 16 November 2026 (UTC):