The problems behind the ideas.

A problem-first view of the deck: the problems people feel today, the agent economy, programmable value, and the patterns that cut across. Ratings are internal, on the same 1 to 5 scale as the idea grid.

Part one

Problems people feel today

Money makes people anxious, so they avoid it

43% of Gen Z avoid their banking app; fintech's top complaint is fear, not features. The bet: lead with opportunity and action, not the number.

Staying informed without being consumed

The feed keeps you scrolling, venues have dead screens, and there is no calm way to keep up or react. The bet: news you do something with, or glance at, not doomscroll.

Connecting with real people is slow, and it exposes you

Starting a group takes forever and costs your number; your presence is static; families split across platforms. The bet: instant, private, no-contact-sharing connection.

Part two

The agent economy

Barely begun, and the biggest opportunity. Today's agents reach consumers as chatbots, not as generative interfaces, and they are barely proactive. That gap is the opportunity.

It has not started: identity, control and payments for agents are all unbuilt at consumer scale, and being early is the bet.

Chatbots, not generative interfaces: agents are still a text box you prompt. Agents that build their own interface and act without being asked is exactly what Agentic Desk does.

Agent-to-agent payments badly need a consumer version: the plumbing (x402, stablecoins) exists, but there is no human-friendly front door.

Part three

Paying and programmable value

Conditional money, group money, per-call machine money, money on a habit: none of it is native to consumers or the web, and stablecoins change the economics. The bet: programmable, instant, stablecoin-settled value made simple.

Part four

Cross-cutting patterns and opportunities

The patterns that show up across many ideas, and where the opportunity sits.

Behaviour change with skin in the game

Dashboards and habit apps only track and shame. Putting something on the line actually moves behaviour, especially when someone else is in your corner.

Proof of real, in a synthetic web

AI-generated identities and content are flooding in. Value accrues to whatever can prove realness: a real human, a real meeting, an agent that is labelled and not impersonating.

Escaping the phone number

The phone number is a broken primitive for identity and contact. Decoupling it, with personas and no number, is a structural moat incumbents cannot copy.

No install, instant, web-first

The app download is the friction that kills adoption. Browser-first, QR-and-link entry, working in one tap for people who have nothing installed.

Ambient, lean-back, and physical screens

Not everything should be a scroll. Glanceable ambient surfaces, typography as interface, and the dead screens in venues as an out-of-home channel with location-relevant promos.

Hyperlocal and the real world

Place, footfall, venues, events and in-person presence are the moat AI cannot fake, and the bridge between digital and physical is underbuilt.

Trust and auditability of AI

As AI does more, people need to trust the output: sources, confidence, cost, a second model checking the first, and honest "not advice".

Value in the data that is not in APIs

The valuable signals (reviews, availability, footfall, satellite) are not in any API; only scraping and agent research reach them. Democratising alt-data is a real edge.

Ephemerality by design

Things that end or fade: it starves trolls, cuts liability, and removes the pressure of a permanent record. A deliberate architecture, not a feature.

Brands as the customer, inside conversations

A recurring revenue pattern: brands and venues pay to live inside chats, rooms and screens, as participants not interruptions.

Regulation as tailwind and constraint

Rules cut both ways: a dated mandate can be a tailwind, and a behavioural rule can be a design constraint that keeps you honest.

Part five

The strategic axes

How to choose between them.

Timing. The agent economy has not started. Some bets are early and huge (the agent stack); others solve problems people feel right now (money and social anxiety, instant rooms).

Meta vs weirdly private. Go big and neutral (Skews, Feed.me, When Then Pay), or go deliberately private and niche (Gazette, Rooms, The Chill Network, Absent Minded).

Ride the rails vs build the rail. For agents and payments the giants build the rails; the edge is the layer on top (Endpoint's routing, The Fixer's control). Do not build the rail.

Works today vs bet on arrival. When Then Pay, Chill Money, Skews and Rooms solve problems now; the agent-economy ideas bet the shift arrives soon.

Consumer-viral vs B2B infrastructure. Viral consumer (Rooms, Skews, Mint, Gazette) versus B2B and infrastructure (KYA, Endpoint, Agentic Desk). Agentic Desk is the exception: B2B, but with a data moat and a finance edge.