Local information is broken
Think about the last time you actually needed something local. Not "a restaurant in San Francisco." Something specific. A barista who just opened a café two blocks from you. A pop-up happening tonight. Someone giving away a couch. A plumber your neighbor just used and would actually recommend. A power outage on your street. The pediatrician three families on your block already trust.
That information exists. Real people know it. It just lives nowhere useful.
Where does the actual, fresh, useful local information about your block live in 2026? Honest answer: it doesn't live anywhere. It's scattered across people's heads, group chats, and 14 different apps that don't talk to each other. That gap is the problem.
The "local" layer of the internet was strip-mined by national platforms that don't actually care about your neighborhood — they care about ad inventory. And the institutions that used to fill the gap are gone.
"What's good near me, right now?"
Pick the smallest, sharpest version of the problem and that's the wedge. Ours is a single question: what's good near me, right now?
That sentence has four words doing a lot of work. Good means trusted by real people, not paid placement. Near means walking distance, not "in the metro area." Right now means open, available, this hour, this evening — not "still in business as of 2023." And me means me specifically — what fits the way I live, where I am, what I actually need today.
Try asking that question of any product you currently use:
- Google: ten blue links and a generative summary that hallucinates whether the place is still open.
- Yelp: a list ranked by ad spend and four-year-old reviews.
- Maps: a pin on a place. No idea if it's any good. No idea if it's actually open.
- Nextdoor: a thread about parking. Nothing about food.
- ChatGPT: a confidently-worded answer based on the internet circa eighteen months ago.
Local businesses have the same problem in reverse. The new café two blocks from you wants to reach the fifty people who would actually walk in today. There is no product on earth that lets them do that for under $500. So they pay Meta to reach 40,000 people who don't live anywhere near them, get no walk-ins, and conclude that "advertising doesn't work."
The wedge is to answer that one question better than anyone else, for one neighborhood at a time, and then for the next, and then the next. Everything else — the feed, the AI, the partner program, the ad platform — is in service of making that single answer faster, fresher, and more honest than any other product on your phone.
A neighbor-posted, distance-first feed
The first thing Manifest does about the broken local layer is the most obvious one: ask the people who actually live somewhere to post what they know. Not crawl. Not scrape. Not license. Post.
That sounds like every other social product and it isn't. Two design choices make it different:
The distance-first part is the load-bearing decision. Every other social platform is built on engagement signals, which means whatever is most outrageous, most argumentative, or most algorithmically reinforcing wins. That works fine for entertainment. It is exactly wrong for local. The thing five miles away might be interesting. The thing two blocks away is useful.
And because the platform is local-first instead of engagement-first, the kind of person who posts is different too. People post the new spot they actually like. They post the lost dog. They post the chair they don't need anymore. They post the contractor who showed up on time. The data layer that emerges underneath is genuinely useful — and it didn't have to be scraped from anyone.
How Manifest works, end to end
The product is intentionally small. Five steps, no taxonomy to learn, no onboarding wizard. Here's the whole thing:
Manifest uses your location to load a real-time feed sorted by distance. Closest first. No account needed to browse, no registration wall, no algorithmic feed. What's happening near you, in the order it's happening near you.
One composer. Selling a bike, hosting a block party, asking for a vet recommendation — same flow. No category selection. No tagging. Photo optional, location auto-attached. The post appears on everyone's feed within your radius immediately.
Not DMs. A visible thread attached to the listing. Multiple interested buyers can see each other's questions and the seller's answers. Neighbors can pile in with recommendations. The social proof and velocity are visible to everyone, which is what private messaging can't do.
"Good matcha near me, open now?" "Anyone know a reliable plumber on the east side?" "What's happening this weekend?" Mani answers from live community posts, partner news articles, OpenStreetMap places, current weather, and the time of day — not from generic web search.
Every post is also a pin. Every city has a map view of what's actually happening — coffee, matcha, coworking, pop-ups, events, services. Open right now, posted by real people, ranked by distance.
Every post helps the person who posted it — and adds a data point to a living, neighborhood-level map of what's happening. Over time the aggregate of those posts is something no scraper could assemble: first-person, timestamped, location-tagged, community-owned local intelligence.
Mani — an AI that actually knows your block
ChatGPT, Gemini, Claude — they're all extraordinary at things that were on the internet eighteen months ago. They are structurally bad at "what's happening near me right now." Trained on web crawls. Generalized across geographies. Stripped of the hyper-local specificity that's the whole point of the question.
Mani is the local-grounded AI that sits on top of the Manifest data layer. It doesn't try to be a better ChatGPT. It tries to be the only product that can honestly answer the wedge question.
What Mani is grounded in
- Live community posts. Real-time, timestamped, location-tagged content from your actual neighbors. The freshest local data on the platform.
- Partner news articles. Reporting from local newsrooms and chambers of commerce that have signed up to power the network — semantically embedded so Mani can pull the relevant article into an answer.
- OpenStreetMap places. Businesses, amenities, transit, parks, and geographic context. Open data, not a Google Maps license.
- Time and weather. Hour of day, day of week, and current conditions are injected into every prompt so Mani knows whether the question is "what's open right now" or "what should I do this weekend."
- Real opening hours. Parsed from OSM so Mani says "open now, closes at 10pm" instead of guessing.
The difference shows up the second you ask it something specific. "Where can I get good matcha near me, open now?" doesn't return ten blue links and a hallucination. It returns a #1 recommendation, says why, references the time of day and weather, and shows what neighbors have actually said about the place.
The next two years of search are being rewritten by generative AI. Whoever owns the data layer that grounds local answers — fresh, first-party, location-tagged — owns the local layer of generative search. That's the real prize, and almost nobody is building for it.
The structural independence is the point. No API key to lose. No platform policy that changes overnight. No case number coming in the mail. The data was created by the community in the first place — so the community gets the answer, not an aggregator that's about to lock the door behind it.
A revenue loop for the people who used to know your neighborhood
Local journalism didn't disappear because nobody wanted local news. It disappeared because the business model — print classifieds and display ads — got vacuumed up by Google and Craigslist, and nothing replaced it. The reporting that used to fill the local layer of the internet has been disappearing one newsroom at a time for two decades.
If we want a healthy local information layer, the people producing the actual reporting have to get paid. Manifest's partner program is built around that.
On the other side of the loop, local businesses get something they've never had: a way to actually reach the fifty people two blocks away who would walk in today. Not the 40,000 people in the metro area Meta will sell them. Real geo-targeting, starting at $10, no agency required. The shop owner, the freelancer, the person hosting a pop-up — they can reach their actual neighbors in five minutes.
Each side of the loop reinforces the other. Publishers' reporting makes the answers better. Better answers attract more readers. More readers create more value for the local businesses that advertise. More businesses generate more revenue back to the publishers. The local data layer pays the people who fill it.
Why this is urgent — the next two years decide it
Important and urgent are different things. The local information layer has been broken for a decade. What's new is that the window to fix it with a community-owned alternative is closing fast.
Three things are happening at once:
- Generative search is replacing the open web. Google's AI Overviews and ChatGPT-style answers are eating the click-through traffic that funded local journalism's last decade. By 2027, "ten blue links" won't be the default surface for most local queries. Whoever owns the data layer those answers are grounded in owns local discovery.
- Local journalism is collapsing on a timer. Over 2,500 US newspapers have closed since 2005, and the rate is accelerating. Without a new revenue model, the institutions that produced trustworthy local information will not exist in five years. The reporting Mani draws on has to be paid for, and it has to be paid for now.
- Big tech is locking down the alternatives. Google v. SerpApi, Reddit v. Perplexity, the unwinding of hiQ v. LinkedIn — the entire data scraping economy is being relitigated in real time. Building on someone else's local data API is a strategy with a pending case number attached.
The window to build a community-owned local data layer — one that doesn't depend on Google's tolerance, doesn't extract from a platform that can sue, and pays the journalists and small businesses on either end — is open right now. In two or three years it won't be. Either it exists, or whoever wins generative search owns local information the way Google owned web search.
Don't scrape. Don't license. Ask the people who actually live somewhere to post what they know, build the AI on top of that, and pay the publishers and businesses that complete the loop. The data isn't ours. It's the community's, shared voluntarily in a context designed for exactly that exchange.
One question: what's good near me, right now? One AI that actually knows. One revenue loop that pays the journalists and small businesses who fill the layer. That's the important problem. The urgency is that it has to exist before the alternatives finish closing.
Manifest is free and available now. Sign up here, ask Mani what's happening near you, or — if you're a local newsroom — join the partner network.