Ideas Radar: 2026-05-23

People want a way to know which accounts and posts on social platforms are genuine humans versus AI agents pushing an agenda. The frustration is specific: feeds are filling with content that is intentionally misleading, and there is no reliable signal for "this was written by a bot for a purpose." A browser-layer or platform-layer authenticity scorer, flagging likely AI-generated posts and coordinated inauthentic accounts in real time, is a real and growing need as the bot-to-human ratio climbs. Whoever nails low-false-positive detection here owns a trust primitive every other platform will want to license.
Source: https://x.com/schwanald/status/2057253980678295627

A complementary need: a clean filter that hides or down-ranks obviously AI-written posts in your feed. The complaint is that low-effort ChatGPT slop is crowding out genuine human writing, and users want a toggle to just not see it. This is the consumer-facing flip side of the authenticity detector, less about catching bad actors and more about curating your own reading experience. A lightweight extension that labels and optionally collapses AI-generated text would find an audience fast.
Source: https://x.com/LumenExTenebris/status/2057480127663964300

There's appetite for vertical prediction markets beyond politics and crypto. Multiple people independently wished for markets on the things they actually follow: the renewal odds of their favorite TV show's next season, and the timing of the next major technological breakthrough. The big platforms optimize for high-volume political and sports events, leaving culture, entertainment, and tech-milestone betting underserved. A prediction market that seeds long-tail, fandom-driven questions could capture an engaged audience the incumbents ignore.
Source: https://x.com/ShazeemiS/status/2057526641874669600

A runner wants an app where you point to exactly where your leg hurts after a run and it tells you what you likely did wrong, form, overtraining, shoe, or recovery. It's a precise consumer health need: turn vague post-run pain into a probable cause and a corrective action, without booking a physio appointment. With phone cameras, a body-map UI, and training data, this is buildable, and the market of injury-prone amateur runners is enormous and motivated. The wedge is being genuinely useful for the most common injuries before expanding.
Source: https://x.com/urwifesbf/status/2057443209706389804

A Reddit user wants a tool that takes an existing video, an SNL skit, in their case, and lets them selectively rewrite and replace just specific lines of dialogue while keeping the rest of the video fully intact. This is distinct from generating video from scratch: the need is surgical dialogue editing and re-dubbing on real footage, matching voice and lip-sync. It's a common content-repurposing pain for marketers and creators who want to adapt a familiar clip to their own context. A tool focused only on "swap these words in this clip, keep everything else" would be immediately useful.
Source: Reddit

A builder is openly asking why no one has built a proper screen-time "lockout" app, a focus tool that genuinely locks you out of distracting apps rather than politely nudging you. The repeated failure of existing screen-time features is that they're trivially bypassed, so the unmet need is real enforcement people can't talk themselves around. Paired with this was a wish for a simpler screen-recording tool. Both point at the same buyer: people trying to reclaim attention and document their work without friction.
Source: https://x.com/DanKulkov/status/2057466484159582551

Two separate posts surfaced demand for relationship apps built around specific values rather than generic swiping. One wants an app for people seeking "covenantal intimacy" with moral and intellectual peers instead of hookup culture; another wants a dedicated, siloed space exclusively for people in open relationships. The common thread is that mainstream dating apps flatten very different intentions into one pool, and niche communities feel poorly served. A values-first or arrangement-first dating product trades total addressable market for much higher fit and retention.
Source: https://x.com/localeria/status/2057276229405843811

A developer wants a way to batch-update the model assigned across all their saved AI commands at once, for example, migrating every command from Gemini 3 Flash to Gemini 3.5 Flash in one action instead of editing each by hand. As people accumulate dozens of saved agent commands and skills, model churn makes per-item editing painful. A small but real workflow gap: bulk model reassignment and version migration for prompt/command libraries. Whoever builds the agent-command manager that handles this cleanly removes a recurring annoyance for power users.
Source: https://x.com/matdegouville/status/2057428384590766404

A recurring meta-observation worth heeding by anyone building: everyone is shipping AI wrappers, almost nobody is building AI that fits into an existing workflow. The gap isn't another chat box, it's tooling that lives where the work already happens and respects the user's current process. It's less a single product idea than a filter to run every idea through: does this slot into a workflow people already have, or does it ask them to adopt a new one?
Source: https://x.com/martin_valchev_/status/2057416875307577432