Stop Building SaaS Clones: Use AI to Build Systems That Solve Scarcity and Corruption

The Question We're Not Asking: What Should AI Actually Be For?

On Reddit, a developer issued a frustrated call to action: can we please stop building endless video game clones and hundreds of millions of redundant SaaS apps, and instead build systems that genuinely free people from scarcity and corruption?

The question cuts to an uncomfortable truth about the current AI wave. As generative models drive the cost of writing code toward zero, the vast majority of new projects continue copying existing business models — yet another to-do list app, yet another subscription tool, yet another reskinned game. The explosive growth in technical capability hasn't widened the range of problems people are actually trying to solve.

The author's position is clear: if we want to avoid what he calls the "Dark Path" — a techno-feudal society where AI owners possess everything and ordinary people toil like 17th-century peasants — then we need to start building fundamentally different kinds of things, right now.

Four "Against-the-Grain" Project Ideas

The author shares several projects he's been thinking through. What they share is an attempt to address institutional or structural problems rather than produce yet another consumer product.

Shower Thoughts: Making Ideas Ownable

The first tool is called "Shower Thoughts" — a place to upload the random ideas that pop into your head, the ones that might be useful or might be worthless, with AI providing a scoring layer. The more interesting part is the technical design: the author proposes using Zero-Knowledge Proof (ZKP) tokens to establish attribution and ownership of ideas.

This touches on a long-standing unsolved problem: creative ideas are notoriously hard to establish ownership over. Who thought of it first? How do you get credit when your idea gets used? Introducing ZKP means you can prove "I possessed this idea at a specific point in time" without exposing the full content of the idea itself — a fundamental redesign of the underlying logic of intellectual property.

A note on Zero-Knowledge Proofs: Zero-Knowledge Proof (ZKP) is a cryptographic protocol that allows one party (the prover) to convince another party (the verifier) that a statement is true, without revealing any information beyond the truth of that statement. In the context of establishing idea ownership, this means a creator can prove to the system "I held this idea at a specific timestamp" without publicly disclosing the idea's contents, generating a tamper-proof timestamped credential. This is similar to blockchain hash attestation, but ZKP goes further — hash attestation still requires eventually revealing the original content for verification, whereas ZKP can theoretically achieve "verify without disclosure." For intellectual property protection, this dramatically lowers the cost of proving prior art while eliminating the risk of the idea being copied upon disclosure. Beyond technical feasibility, however, such systems still face real-world challenges around legal recognition and on-chain data persistence.

Civil Good: A Mutual Aid Matching Platform

The second project connects "people who need help" with "people who have spare time and the skills to help." The author openly acknowledges it still needs safety mechanisms — which is precisely the hardest part of any well-intentioned platform like this: how do you prevent abuse? How do you build trust? How do you protect physical safety when people meet in person? The idea itself isn't difficult; the hard part is building a system that can't be weaponized.

Invoice Netting: Doing "Subtraction" on Business Debt

The third idea has the most immediate practical value. Imagine 50 companies all owe each other money, forming a giant cycle of capital flow. The traditional approach is for every company to actually pay out and then collect in real cash. The logic of netting (or equalisation) is to first cancel out all the circular "owe" and "owed" relationships, then only settle the final net amounts.

This is actually the real-world concept of multilateral netting, but lightweight tools aimed at small and medium businesses remain scarce. It could significantly reduce cash flow pressure and decrease dependence on short-term financing — a simple idea that could have a meaningful impact on operational efficiency.

A note on multilateral netting: Multilateral netting already has mature applications in traditional finance, primarily in interbank clearing (such as SWIFT's net batch settlement) and internal treasury pooling within large corporate groups. Its core mathematical logic is to identify and eliminate debt cycles in a directed graph, compressing O(n²) bilateral payments down to O(n) net payments. A simplified example: A owes B 100, B owes C 80, C owes A 60. Through netting, only two transactions are needed — A pays B 40, and B pays C 20 — rather than three independent transfers totaling 240 in capital movement. The practical value for SMEs is particularly significant: small businesses typically lack ample credit facilities to smooth their receivables cycles, and capital tied up in unpaid invoices directly threatens operational survival. Lightweight netting tools for SMEs are genuinely scarce today, with the main obstacles being trust coordination among participants, legal enforceability, and standardization of multi-party data access.

New Governance Frameworks: AI as Evaluator or as Designer?

Finally, the author has written several pieces on new governance frameworks intended to replace what he sees as corrupt political systems. He distinguishes between two roles for AI: one where AI acts as a partial evaluator, research assistant, or modeling tool; and one more radical — where AI designs the system itself, which humans then independently operate by hand.

This distinction is worth dwelling on. It reflects two fundamentally different attitudes toward AI's involvement in power structures: one that preserves ultimate human decision-making authority, with AI as a tool; and another that delegates institutional design to AI, but deliberately keeps execution in human hands to prevent algorithmic rule.

After Democratizing Technical Skill, What's the Real Bottleneck?

The author himself identifies a key contradiction: these projects haven't been publicly released yet, and "anyone can basically vibe-code an identical version instantly with Opus and Astra."

This sentence reveals a profound shift. Once coding ability has been democratized by large models, "can we build it" is no longer the barrier. The real bottleneck has moved elsewhere: choosing what to build, designing incentive mechanisms, addressing real-world safety and trust problems, and keeping a well-intentioned system from being captured by capital or power.

In other words, what's scarce is no longer the hands that write the code, but the minds that ask the right questions — and the institutional imagination to translate social problems into systems that can actually run.

A note on "vibe coding": "Vibe coding" is an informal term that's become popular in developer communities, referring to the practice of using large language models (such as Claude Opus, Google Astra, etc.) to rapidly generate working code through highly natural-language-driven prompting, where the developer plays more the role of requirements describer and acceptance tester than line-by-line author. The spread of this paradigm has compressed prototype development timelines from weeks to hours, significantly lowering the barrier to entry. The author uses the term to emphasize that once a product idea is made public, almost anyone with initiative can immediately replicate its technical implementation. This fundamentally changes where "first-mover advantage" comes from — in an era when code can be generated instantly, moats no longer come from technical execution ability, but from deep understanding of the problem, accumulated community trust, and the ability to design incentive structures that resist abuse.

A Call Worth Taking Seriously

Setting aside the idealistic, even utopian elements, the core argument of this post holds up: the democratization of technical capability should have opened a window for solving large-scale social problems, but the gravitational pull of capital keeps directing the vast majority of creative energy toward whatever is easiest to monetize.

The directions the post sketches out — idea ownership, mutual aid matching, business invoice netting, AI-assisted governance — may not all be mature, but they represent a class of problem spaces that are severely underexplored. In an era when anyone can rapidly build applications, the real differentiator may not be the tech stack, but whether you're willing to point your compute at problems that are hard to build, hard to profit from, and genuinely important.