Skip to content
ARCEDE

Article

Working with philanthropy

Here's what I keep coming back to. Most of the people who can deploy frontier systems end up working on commercial software, consumer markets, or the next platform layer, because that's where the most capital is being deployed. The organizations you support, doing consequential public-benefit work, can't match those rates. They shouldn't have to alone.

Without funders like you deliberately closing that gap, the capabilities of intelligent tools keep flowing toward commercial markets and away from public-benefit work. Closing it is exactly the kind of thing your capital is positioned to do, and the work I'm most interested in doing alongside you.

On leverage

Conversations about intelligent tools in commercial work usually settle into labor substitution once demand is captured: output multiplies, the team contracts, profit lifts. Roughly this chart: the modeled case where the market is already near saturation and there is no obvious new demand to open.

Deploying frontier systems in commercial work: leveraged output, headcount under pressure.A chart showing three curves over a twelve-month horizon with a horizontal baseline at the middle of the canvas. Above baseline, a terracotta curve labeled deployed frontier systems rises moderately, and a sage curve labeled output rises substantially more: the leveraged commercial output. Below baseline, a slate-blue curve labeled headcount falls in a saturating decline, stabilizing at roughly 60 percent of its starting level by year-end. The visual claim: in commercial work, leverage is captured by reducing the team rather than redirecting hours to mission. This is the labor- substitution dynamic the philanthropy framing on this page argues against.COMMERCIALoutputdeployed frontier systemsheadcount
Output multiplies, the team contracts, profit lifts. The standard commercial picture.

The math is different in the work your capital supports. The unit of value isn't profit; it's impact. These systems take on the assembly cost of a mission-driven team: e.g., board prep, grantee report digestion, monthly reporting. Those hours don't come out of headcount. They come back to the team, redirected to other work to further impact. Leverage doesn't replace your team, it allows them to focus on higher-order items and impact.

Deploying frontier systems in philanthropy: leverage on impact, team preserved.A chart showing three curves over a twelve-month horizon, all starting at the same dimensionless baseline. A sage curve labeled impact rises sharply in the first two months, re-accelerates around month six, and ends near three and a half times the baseline. A terracotta curve labeled deployed frontier systems rises modestly to about one and a half times by year-end, representing the input being put to work. The widening space between the two curves is the leverage of the underlying technology compounding while the systems are running. A slate-blue line labeled headcount runs flat at the baseline, representing labor preserved. The visual claim: modest deployment, leveraged impact, team unchanged.PHILANTHROPY2×3×4×1×impact≈ 3.6× AT MONTH 12deployed frontier systemsheadcountOVER 12 MONTHS →
Modest deployment, leveraged impact. Three and a half times over a year. The team stays; the hours go to expanding impact.

Notes & sources. Y-axis is throughput multiplier from the team’s current baseline: 1× = today’s throughput, 2× = double, 3× = triple, 4× = quadruple. The sage line is impact, rising to ≈ 3.6× at year-end. The terracotta line is deployed frontier systems, the operational input being put to work, rising modestly to ≈ 1.6×. The widening space between them is the leverage of the underlying technology compounding while the systems are running. The faint inner band on the impact line is the 80% confidence interval from a 200-sample parameter Monte Carlo; where the optimistic tail breaks past the 4× ceiling near year-end the band is clipped at the chart top rather than rescaled. Curves modeled by the Arcede Frontier-Impact Model v0.3, which decomposes impact into three sequential paradigm stages (workflow automation, supervised reasoning, autonomous- class systems) with a time-varying frontier doubling rate. Frontier rate calibrated against METR Time Horizon 1.1 (May 2026), Epoch AI capability progress data, and the Stanford AI Index 2026. Other parameters are operational defaults; calibration data accumulates per engagement.

The value that an engagement with Arcede creates may be profound due to the leverage it creates for driving impact. Intelligent tools are like any tool a person uses and everything that comes with that. Frontier systems fundamentally rewrite productive capacity.

What this usually looks like

A few examples of what this kind of work looks like in practice.

I work directly with the organizations you already support: workshops, training, or forward-deployed engineering across multiple grantees, coordinated rather than scattered. I embed with your grantees to understand workflows, with you underwriting the work so they don't have to find an engagement budget on their own, and so they can use this technology to scale or improve the quality of their impact.

Or the support is more operational: underwriting my ability to keep doing this kind of work at the rates the sector can actually sustain, and to expand the practice around it. Variants and combinations of these are directionally where it makes sense for us to land.

Mechanisms

The right mechanism depends on your structure. Direct grants via expenditure responsibility, grants via fiscal sponsorship, program-related investments, direct contracts for service, or hybrids of those are helpful.

Why this matters

As the technologies of the intelligence age rapidly accelerate, there will be big winners, and there will be people and institutions left behind by arbitrary gatekeeping. I think that is structurally unfair. The small group of people who understand these emerging intelligent tools and can build systems around them will wield tremendous leverage.

Part of my mission is to disperse that leverage in pragmatic ways to partners, potentially such as yourself, to drive positive societal impact and address some of the grand challenges we all face.

News

More news.

All news
  1. Article

    AI Oversight and Verification Saturation

    AI raises how much an institution produces. It does not raise the hours to check it.

    Read the article
  2. Papers

    Payment, identity and communication

    Three new papers in The Agent Internet series, on what an agent needs beyond access: the ability to move money, to prove it belongs, and to talk to people.

    Read the papers
  3. Apolitical

    LOVS: turning public disease outbreak data into spatial maps, forecasts, and searchable intelligence

    A case study on the Apolitical Navigator describing how public outbreak reporting is consolidated into spatial maps, forecasts, and searchable evidence.

    Read at Apolitical
    (opens in a new tab)