Seven Gaps Where People Asked for a Recommendation, Not an App

Some of the strongest demand on our board is not a request for software. It is a request for an answer. As of 2026-10-28, GapMine tracks 1,196 open gaps, and we pulled seven of them where every quote is a person asking which part to buy or which setting to give up. We think this is the most overlooked shape of demand we hold. It reads as unbuildable until you see what people are stuck on: nobody will tell them what to pick.

The constraint comes first, the product second

The clearest case is Local LLM Model Selection, with 19 people and 15 rivals shipping. One of them wrote: "I don't need more VRAM, I need to know what to sacrifice: lower model quant, quantize KV cache, reduce context length, or sacrifice inference speed." That is a four-way tradeoff with a known hardware ceiling. Benchmarks exist. A tool that turns the ceiling into a ranked set of sacrifices does not.

Easy Homelab Setup drew 7 people and 12 rivals shipping. The ask is budget-shaped: "I don't need an epyc server cpu to be super expensive, I would like to have an epyc server cpu, like a 7502 in it." They already know the part number. What they want is a path to it at their price.

Best Disk Storage Options has 5 people and 30 rivals shipping, the heaviest supply in this group. Supply is not the problem. "I don't need a huge setup, my sff can only host one internal drive" is a filter nobody offers, and thirty shipping products make the filtering harder, not easier.

Homelab Nas Cpu shows 5 people and 17 rivals shipping, with the same pattern: a chosen chip, an unchosen everything else.

Mac Mini Microphone Upgrade sits at 5 people and 13 rivals shipping. The request names the machine, the use case, and the software: "I need a microphone recommendation for my Mac mini M4 setup. I primarily use it for FaceTime and streaming games with OBS." Three constraints, no match engine.

Accurate Outdoor Temp Monitoring has 5 people and 8 rivals shipping, and the quote is the whole gap in one line: "Which outdoor Zigbee temperature sensor would you recommend?"

Then there is PCIe x16 to dual x8 adapter, with 5 people and 3 rivals shipping. "I am slowly going insane. Either I have lost my ability to browse the internet or there are no PCIE 5.0 x16 to x8x8 splitters." This one may genuinely be a missing part rather than a missing recommendation, and it is the thinnest supply in the set.

The pattern

Supply here is thick, and that is normal for our board. As of 2026-10-28, of 1,196 open gaps, 884 cards sit in the crowded supply bucket of ten rivals or more and only 47 sit in the thin bucket of one to three. If you are hunting an empty market, these are not it. The opening is in the selection layer on top of a full market, which is a different business with a different distribution plan. It is close to what we saw in gaps where people asked for fewer features, and it is why the truly underserved niches are rarer than the listicles suggest.

The practical read: a configurator that takes a hard constraint and returns a ranked pick, with the tradeoff spelled out, has demand in at least seven places on our board. The hardware ones are affiliate-shaped. The LLM one is not.

Honest limits

These counts are distinct public authors, not market size. A card backed by a small handful of distinct authors can mean a quiet niche or a thread that happened to surface. Our rival scan has a recall ceiling: it finds shipping products, it does not prove a product is absent, so a low number means we found few, not that few exist. Several of these cards are hardware adjacent, and software founders may find the build is a content and data problem instead of a code problem. Dates on the cards are when the gap was logged, not when demand peaked. The grouping of these seven is our judgment about what people meant, not a field in the data.

All numbers from GapMine's live dataset, same pool the site renders, as of 2026-10-28.