What if you could replace corporate budgets with decentralized prediction markets? What if strategic work could be directly translated into live code by an AI, verified by a human, and paid for automatically via a crypto-style bounty, all in a matter of hours?
This isn't a thought experiment; Jonathan Spector and I actually ran it last week.
At GSD at Work LLC, we're exploring a new, emerging paradigm for work: Jonathan was brave enough to share his expertise with me, and then dive headfirst into a complex AI system and start pushing changes to our live website, was willing to accept payment through a programmable bounty platform, and stake reputation (as well as play money, for now) via a public prediction market.
The result wasn't just a new section on the GSD website; it was a functioning prototype of a radically different way to organize, incentivize, and coordinate knowledge workers.
This is the story of how it happened, and what it might tell you about the future of work.
The Setup: A Fusion of Strategy and Execution
The experiment began with a classic business problem: aligning our company's messaging with a newly-defined Ideal Customer Profile. Our company, GSD, specializes in AI-driven business transformation. Jonathan Spector, an executive-level strategy consultant with decades of experience in strategic alignment, had been working with us to sharpen our focus. Through a series of workshops, we'd identified private equity-backed companies as a key target. They have concrete EBITDA targets, time constraints, and a clear need for the operational efficiencies AI can deliver-and we've been lucky enough to serve them more and more over the past year.
The challenge was to translate this high-level strategy into tangible changes on our website that would actually speak to the ICP. The old way would involve a linear, multi-step (dare I say, "waterfall-esque") process: Jonathan would deliver a strategy deck, which would then go to a marketing team, then to a copywriter, then to a developer, with feedback loops and delays at every stage.
We decided to try a different path-one that compressed this entire workflow into a single, collaborative, AI-mediated session. Jonathan's role was not to hand off a strategy, but to participate directly in its execution.
The Experiment: From "Yak Shaving" to an Automated Cascade
The plan was ambitious: Jonathan, self-described as "not a coder," would make a direct code contribution to the GSD website repository on GitHub. This contribution would trigger an automated bounty payment through the platform Boss.dev and simultaneously resolve a public prediction market we had created for the task on Manifold.
The process began, as technical endeavors often do, with a period of frustrating but necessary setup work-a process known in developer circles as "yak shaving." We spent the first part of our session navigating the intricacies of command-line tools, GitHub authentication, and Windows-specific configurations while trying to install Claude Code (which I love) on Jonathan's machine. For all the talk of vibe coding and software engineering expertise being obviated, friction still exists when transforming business expertise into production code.
But then came the pivot. Instead of forcing a complex local setup, we shifted to a cloud-based AI coding environment (ChatGPT's Codex) that was already connected to our GitHub repository. This is where the magic happened.
Jonathan didn't write a line of code. Instead, he dictated a high-level prompt to the AI, feeding it the raw material from our strategic discussions and a document he'd prepared summarizing our new ICP messaging.
His prompt was simple and direct: "I want you to make changes to the GSD at work website based on some of the work that Christian and I have done... as driven or indicated by the document that I am going to reference here."
The AI took this strategic directive and, within minutes, generated four distinct versions of the website, each with dozens of specific code changes. It had rewritten headlines, added new sections tailored to PE firms, and embedded the core ROI messaging we had discussed. Jonathan reviewed the options, selected the one that best captured our intent, and with a few clicks, submitted a "pull request"-a formal proposal to merge his AI-generated changes into the live website.
I reviewed the request, merged it, and then the cascade began.
The website updated instantly. The new messaging focused on PE-backed firms was live.
The bounty was paid. The Boss.dev platform detected the merged pull request and automatically executed a payment to Jonathan's connected account.
The prediction market resolved. The Manifold market we'd created-"Will GSD's website be updated with Jonathan's ICP/messaging recommendations by August 7th?"-automatically resolved to YES, triggered by a tag on the completed GitHub issue.
The entire end-to-end process-from strategic prompt to live, paid-for execution-was complete. My reaction, caught by Fireflies.ai, was one of genuine astonishment: "Holy smokes. It all worked... I'm glad this is being recorded because this is like... I think this is a historic moment, to be honest."
The So What: A New Operating System for Work
This experiment was more than just a clever chain of automations. It was a tangible demonstration of three core principles that could underpin a new operating system for work.
1. From Budgets to Prediction Markets: Decentralized Capital Allocation The prediction market wasn't just a gimmick. It represents a shift from top-down budgeting to bottoms-up, decentralized capital formation. Instead of a manager allocating a budget, a project owner can state an objective and their willingness to pay for it. Anyone who believes they can contribute to that outcome can "bet" on its success by buying "YES" shares. As I explained to Jonathan, this model inverts traditional incentives: "It actually makes 'insider trading' a good thing because I want to attract people who feel... confident that they can help achieve this outcome." It's a mechanism for coordinating work and allocating capital based on demonstrated confidence and accrued reputation.
2. From Handoffs to Human-AI Collaboration Jonathan, the expert, became the orchestrator of an AI developer. He didn't need to know the syntax of React or TypeScript; he needed to know how to articulate a clear strategic goal. The AI handled the tactical execution, and the human handled the final qualitative review and verification of a good work product. This redefines roles, allowing strategic leaders to become direct, value-creating contributors without getting lost in technical minutiae.
3. From Trust-Me to Trustless Verification The entire system was built on verifiable, automated trust. The completion of the task was not a subjective judgment but a cryptographically-signed event on GitHub. This event triggered both the payment and the market resolution. There was no need for invoices, approvals, or payment processing departments. The work and its reward were inextricably linked in an automated, transparent ledger.
This experiment wasn't seamless, and the "yak shaving" highlights the work still to be done in making these tools accessible. But it proved that the model is viable. It showed that we can build systems that are not only more efficient but also more dynamic, transparent, and meritocratic.
For Jonathan, it was quite the learning experience, moving him from strategist to a direct participant in this new economy of work. For GSD, it was a validation of our core thesis: that the future of work isn't just about using AI as a tool, but about redesigning the very structures through which we collaborate and create value. We didn't just talk about the future; for a few hours on a Wednesday afternoon, we lived it.
Want to accelerate your AI-era retooling journey and even participate in our next experiment?
Learn our methodologies, and when you're ready - apply to become an Associate.