Ideas Radar: 2026-05-11

A way to short memecoins natively, on-chain, without going through some janky perp DEX where the liquidity dies before you can close. The pitch is obvious — every memecoin cycle creates billions in obviously-going-to-zero positions, and the entire trader base has to sit on the sidelines or try to find a CEX that lists the right perp at the right size. The product is "Polymarket but for memecoin decay" — fixed-duration short contracts on tokens above a market cap floor, settled in stables. Risk-managed pool. Capital-efficient. Pure infrastructure play with day-one traction.
Source: https://x.com/Romeo_Onchain/status/2053108506371199482

Retail investor access to private AI company equity — specifically the ones every public investor wants to own (Anthropic, OpenAI, Mistral, xAI). Fractional shares, secondary-market structure, KYC-gated SPV access. The accredited-only path through SPVs already exists but is locked off from the 99% of retail who actually want this exposure. Crypto solved the same problem for early-stage protocols with token sales — the regulated-equity version is the obvious next move. Whoever cracks the legal wrapper first owns the category.
Source: https://x.com/Dannytheboy94/status/2052982068083151305

A user setting in ChatGPT to manually pick which model handles your turn instead of auto-routing. Power users hate that GPT silently swaps brains mid-conversation depending on inferred difficulty — they're losing the consistency a single model gives. The feature ask is one toggle: route to my choice this session, with auto as the default. Replies frame it as a question of trust and predictability for serious work. This is a feature OpenAI could ship in a day but probably won't because routing is their inference cost-optimization play. So the third-party wrapper opportunity exists.
Source: https://x.com/hopes_revenge/status/2053157687735750920

Long-term car parking concierge for people traveling 2-6 months. Not just a parking space — a service that drives your car around the block every few weeks to keep the battery alive, the seals from drying out, the brakes from rusting. The closest thing today is asking a friend or paying a neighbor. The market is anyone who travels for work, sabbatical, military, or extended family caregiving. Bundled with insurance for the in-driveway period, this is a $50-150/month subscription per car in any metro with rich car owners.
Source: Reddit

A US-based tourbillon watchmaker capable of building and maintaining $250-300k mechanical watches. The current map of who can do this in America is essentially empty — every serious tourbillon repair goes to Switzerland or Germany, sometimes Japan. The market is small but the per-unit margin is enormous. Pairs naturally with a training-and-apprenticeship pipeline — there are exactly zero accredited US tourbillon training programs. A boutique manufactory plus a one-year residency program could anchor an entire micro-industry.
Source: https://x.com/MileHighAva/status/2053123968928985509

A fintech platform — quasi-bank, money-market, or stablecoin-backed deposit account — that publishes real-time, fully transparent balance sheet data plus private-market deposit insurance. The trigger today is that bank failures still happen and FDIC has a $250k cap; high-net-worth depositors have no clean way to verify the institution holding their money is actually solvent. The product is verifiable reserves plus a private insurance wrapper. Existing crypto platforms have nibbled at this but none have a real proof-of-reserves + private DI structure. Big TAM.
Source: https://x.com/Stef_McConnell/status/2053231384836141161

A judicial accountability database. Pulls public court records, links repeat offenders to the judges who oversaw their cases, exposes patterns — lenient sentencing, suppressed bail, plea deals that didn't stick. This is civic tech with a clear customer in journalism, prosecutors' offices, voters, and victim-rights groups. The raw data is mostly public but scattered across hundreds of county sites. The wedge is one well-designed state or federal district, expand from there. Equal parts data engineering and UX problem.
Source: https://x.com/Robertstinker/status/2053104218978877540

Mach-E (Ford EV) owners are accidentally writing a product roadmap. The most-quoted gaps from one detailed two-week review: granular EQ control over the B&O sound system, a Now Playing widget for the digital dash, customizable dash info combinations, stronger regen-braking option for true one-pedal driving (the current setting is too weak and tied to drive mode), an "MPG off, percentage only" battery display toggle, and a true FSD-equivalent driver-assist feature. Ford is shipping none of these. An aftermarket "Mach-E Plus" software layer — or a Tesla-Plaid-style modding scene — has real demand if the regulatory wrapper works out.
Source: Reddit

An anime recommendation engine that actually understands preference instead of "people who watched X also watched Y." The volunteer offering to build this today is just one of dozens of people on the timeline frustrated with the current crop of recs. The win is taste-graph plus episode-level metadata plus social signal (which arcs your friends rewatched, where they bailed). MyAnimeList has the data but not the recommendation engine. AniList has the engine but not the trust. The opening is a third-party that does both, monetized via Crunchyroll/HiDive affiliate revenue.
Source: https://x.com/mayorxbt/status/2053227050735947966

Reverse Y Combinator: an accelerator specifically for artists who want to become entrepreneurs. The current ecosystem assumes the founder is technical and is teaching them business; this would assume the founder is creative and teach them ops, ARR, hiring, distribution. Successful examples already exist informally — Beeple, Olafur Eliasson, the Stripe Press model — but no structured cohort, no shared playbook, no peer cohort cap table. The natural funding angle is collector LPs who already buy the work and want their favorite artists to scale.
Source: https://x.com/akashgeny/status/2053204392489161175

An "auto-block anyone who blocks you" feature for X. The complaint is that people use block as a weapon — block to remove dissenting replies from a thread without consequence — and the target has no symmetric response. The feature is a one-line account setting. X almost certainly won't ship it because it nukes engagement, but a third-party Chrome extension or a dedicated viewer client could. Adjacent feature: "see who blocked you" — Twitter Lite clones in the past have tried this and bumped against API limits, but the appetite is huge.
Source: https://x.com/iamKierraD/status/2053184441611141576

A D&D companion app that tracks modifiable stats (Fatigue, Discovery, custom condition tracks) that actually affect rolls. D&D Beyond and Roll20 are stuck at the official stat block. DMs run the most interesting campaigns with custom mechanics, and those mechanics live on paper and Google Sheets because no tool models them. The product is "homebrew stat engine" — a layer that sits on top of D&D Beyond and exposes a clean API for custom counters, modifiers, and roll-time triggers. The market is small but extremely high-LTV — DMs run multi-year campaigns with the same tooling.
Source: https://x.com/TaleMasterTOV/status/2053245119009309069