Ideas Radar: 2026-04-20
Today's need-signal is weirdly dev-facing — people keep asking for AI-ready databases and AI-filtered social tools. There's also a stubborn friend-finding app gap nobody has solved, a janky X index API data goldmine sitting unused, and a very 2026 ask for a janitor that scrubs your political footprint off your own timeline before a hiring manager reads it.
#1
A reactive database where queries are TypeScript functions — pitch is specifically that LLMs would do very well with it. The bet is that the query language of the future isn't SQL or GraphQL but code agents already write, and that reactivity gives you a live feedback loop every time the dataset changes. Whoever ships the right primitives here (observable query functions, incremental compute, type-safe subscriptions) becomes the Supabase of the agent era.
Source: https://x.com/jamesacowling/status/2045335994803859620
Source: https://x.com/jamesacowling/status/2045335994803859620
#2
A self-censorship tool that scans your social timeline and scrubs or flags posts and comments on political/sensitive topics that a hiring manager might use to reject you. Very 2026: hiring loops already run AI on candidate public footprints, and people know it. Target users: anyone job-hunting in politically charged regions. Same engine extends trivially to dating-app audits or college admissions grooming.
Source: https://x.com/capjmk/status/2045175706007601354
Source: https://x.com/capjmk/status/2045175706007601354
#3
A mass-unsubscribe service for physical direct mail — not email. Inbox unsubscribe is solved; paper spam is not, because opt-outs are scattered across DMA lists, catalog company forms, and county property records. A real product would handle the paperwork per country/state and sell on time saved plus trees saved. Works as SaaS subscription or per-scrubbed-address pricing.
Source: https://x.com/almostcmb/status/2045158649320927433
Source: https://x.com/almostcmb/status/2045158649320927433
#4
A Tinder-style app purely for platonic local friend-making — explicitly not social media, explicitly not dating. Core pain is the gap between "Instagram followers" and "people you actually see this week." Bumble BFF exists but users keep asking for something not attached to a dating company. Execution is the hard part (moderation, female safety, fake profiles), but pull is real.
Source: https://x.com/brknmirrorball/status/2045312003263135831
Source: https://x.com/brknmirrorball/status/2045312003263135831
#5
A lightweight competitor analysis tool built on top of X's index API. The observation: the index API is a goldmine sitting unused while everyone pays for bloated enterprise SaaS that's essentially a wrapper. A $49/mo SMB product surfacing competitor mentions, share of voice, ad spend signals, and engagement delta would find customers the day it launches.
Source: https://x.com/blanplan/status/2045484535878672464
Source: https://x.com/blanplan/status/2045484535878672464
#6
Real-world force/tactile and proprioception data infrastructure for training humanoid robots. The specific claim: simulation and internet video can't capture the physical data humanoids need, and nobody is building the factories to produce this data. This is the picks-and-shovels layer of the humanoid era — whoever owns the sensor fleet plus the annotation pipeline becomes the Scale AI of physical intelligence.
Source: https://x.com/rikit_rathi/status/2045478935345987854
Source: https://x.com/rikit_rathi/status/2045478935345987854
📡 Eco Products Radar
Eco Products Radar
X index API — repeatedly named as an untapped data source; obvious wedge for competitor-analysis products.
Claude Code / LLM-first tooling — referenced as the design constraint for the reactive TypeScript DB idea.
X index API — repeatedly named as an untapped data source; obvious wedge for competitor-analysis products.
Claude Code / LLM-first tooling — referenced as the design constraint for the reactive TypeScript DB idea.
Comments