Partner with AI. The future of business is being built where creativity, technology, and human connection meet.

VirtualRoad's mission: empower creators and small businesses to harness modern technology in practical, ethical, and transformative ways.

VirtualRoad helps businesses and creators find, choose, secure, and deploy digital solutions with confidence.

www.virtualroad.ai - (781)913-0603

June 1 – June 11, 2026

This Week's Big Idea:

Three people who agree on almost nothing — Donald Trump, Bernie Sanders, and Sam Altman — just landed in roughly the same place on one question: should the American public own a piece of the AI boom? When the political poles converge, pay attention. Something structural is moving.

We're also digging into a concept that's been sitting with me from a recent AI Daily Brief episode: the shift from prompting to building loops. The single highest-leverage skill of the next 18 months isn't writing a better prompt — it's designing the cycle that runs without you. More below.

This week the market is still digesting the largest set of compute deals ever disclosed. The industry has fully pivoted from "Generative" (creating) to "Agentic" (executing) — and the two stories defining this week, public ownership and private infrastructure, are really the same story viewed from opposite ends: who owns AI, and who powers it.

National Intelligence: "The Concept" - Americans as AI Shareholders

A sentence I didn't expect to write in 2026: Trump, Sanders, and Altman are roughly aligned on something.

  • The pitch. Aboard Air Force One, Trump floated "concepts where pieces could be given to the American public, where the American public essentially becomes a partner with the companies." He called it "very interesting" and said he's meeting with AI executives in the near future. Bloomberg

  • The mechanism. It's genuinely novel. Rather than selling equity, OpenAI would donate shares to the federal government — a structure that avoids any direct cash outlay from taxpayers. Those shares would seed a "Public Wealth Fund," with proceeds that "could be distributed directly to citizens, allowing more people to participate directly in the upside of AI-driven growth, regardless of their starting wealth or access to capital." Think AI dividends — some versions route them to children, others toward paying down the national debt. OpenTools.aiTechCrunch

  • The strange convergence. Trump, Sanders, and Altman have all signaled support for some form of public equity — though they disagree sharply on how much. Sanders proposed a one-time 50% tax that OpenAI, Anthropic, and xAI would pay in the form of stock. On the raw economics, the two camps aren't as far apart as you'd expect. OpenTools.aiTechCrunch

Why this hits home: this is the VirtualRoad thesis scaled to a nation — the belief that AI's upside shouldn't pool at the very top, and that the people doing the work should share in the value the tools create. Watching that idea jump from a North Shore consultancy to Air Force One is surreal.

Now the cynicism, because you've earned it. "The American people become partners" is a warm phrase doing a lot of quiet work. Sanders' bill would also grant the federal government voting shares and equal board representation at each targeted company. As one legal scholar put it, the hard question is how to design systems that prevent concentrations of power — in private or public hands. A central government holding both equity and board votes in the most powerful companies on earth is not an obviously safe outcome. Partnership for the public, or leverage for whoever holds the office? Watch closely. MLQOpenTools.ai

The Infrastructure Land Grab

While Washington debates who owns AI, the people who power it made a jaw-dropper.

  • The deal. SpaceX signed a $920M-per-month contract with Google, running October 2026 through June 2029 — potentially ~$30B total — giving Google access to roughly 110,000 Nvidia AI chips. The Decoder

  • Where it's fulfilled. At data centers tied to xAI, the AI company SpaceX absorbed in an all-stock merger earlier this year. SpaceX is effectively a GPU landlord — providing compute capacity only. Tech TimesMLQ

  • The bigger one. Anthropic is paying SpaceX $1.25B per month through 2029 to rent the Colossus 1 data center near Memphis. Google's deal covers about half that compute. Yahoo Finance

The takeaway: ==when a company with Google's resources finds it cheaper to rent 110,000 GPUs than wait for its own, the AI race has officially become an infrastructure race — decided by who can secure chips and power at scale.==== Nvidia keeps printing. SK Group and the supply chain ride the wave. Behind every clean chatbot answer sits a vast, expensive, contested machine.== Tech Times

The Field Report: Who's Using AI Well

Four companies are modeling the right posture. None treat AI like a fancy autocomplete.

  • KPMG — treats AI as a reasoning partner. You collaborate with it, you don't command it. The mindset is the whole game.

  • Scrunch — uses AI to research how your business shows up to other AI. (That's the CITED thesis: your visibility inside AI answers is the new SEO.)

  • Section — coaches employees on rolling out value, not tools. The tool is easy; the behavior change is the work.

  • OutSystems — builds governed, agentic systems for the enterprise: AI agents running a multitude of tasks inside guardrails.

The through-line: the advantage gap is widening between power users and casual users. Knowledge workers who use agents to solve problems and create opportunities are pulling away from everyone still typing one prompt and calling it a day.

The Loop Economy: Stop Prompting, Start Orchestrating

Here's the most useful thing I can hand you this week.

Most people still prompt their tools — one shot, one answer, repeat. The next tier has stopped prompting and started building loops: designed cycles where an agent works, checks itself, and iterates against your real workflow.

Know this before you go all-in:

  • Loops burn tokens — roughly 10x a single prompt. Not a bug; it's the cost of autonomy.

  • Loops have error rates. An agent running unsupervised compounds its own mistakes. Loop design is what keeps it on the rails.

  • The win: a well-designed loop that fits your workflow delivers real, repeatable value. It's a business case, not a toy.

The advanced behavior pattern to internalize: don't prompt your agent — design the loop that prompts it.

On the tooling front: OpenAI's Codex has quietly become its most successful product, and the rumored "SuperApp" looks more like a company reorg than a new toy. The terminal CLIs, the IDE agents, Claude on the desktop — all converging on one idea: the loop is the unit of work now, not the prompt.

The Pro Playbook: Moving from Prompting to Orchestration

Phase

Action for SMBs

Action for Creators

Audit

Map the one workflow you repeat most every week.

Identify the repetitive research/edit task eating your hours.

Architect

Design a single agent loop around it — work, check, iterate.

Build a "reasoning partner" prompt chain before automating.

Deploy

Run it, measure token burn vs. value, then expand.

Ship one loop, watch the error rate, refine the design.

To do:

Pick the single workflow you repeat most. Map it. Then build one agent loop for it — even a clumsy one. You'll learn more from that than from a month of newsletters like this one.

Forward this to one person who's still prompting like it's 2024.

— Andrew

Keep Reading

© 2026 Virtualroad Newsletter.
beehiivPowered by beehiiv