DHM Ventures: Convex Exposure to What Comes Next

DHM Ventures is an AI-native venture lab. We build technology companies and back founders around a simple premise: as AI makes code, content, analysis, and parts of management dramatically cheaper, value migrates toward what remains scarce — distribution, trust, proprietary data, taste, relationships, and position.

We spent several years rebuilding our operating model around that change. Over roughly the past year, we put the model to work across eight companies. They are not eight independent startups assembled under one holding company. They form a technology tree: companies share infrastructure, tools, operating knowledge, and increasingly product capabilities.

The investment question is whether DHM can use AI to lower the cost and time required to explore technological opportunities, then concentrate capital behind the relatively small number that produce evidence of durable scarcity and market demand. We call the governing metric Time to Future: how long it takes a possible future to become an operating one.

DisclosureThis material is being made available to accredited investors only, as defined under SEC Rule 501(a). Information presented does not constitute an offer to sell or a solicitation of an offer to buy securities; any such offer will be made only through definitive offering documents.

1. The Operating Model

DHM does not treat AI as a productivity tool grafted onto a conventional company. We decompose company-building into work AI has made inexpensive and work that remains scarce.

Strategy analysis, financial modeling, technical evaluation, software production, market research, positioning, operational coordination, and parts of go-to-market can increasingly be performed by AI systems at low marginal cost. DHM's operating stack therefore includes four persistent executive functions: CEO, CFO, CTO, and CMO.

The point is not to pretend that software has become a human executive. It is to preserve multiple functional perspectives without requiring a traditionally staffed management organization. The CFO can frame a product decision as a capital-allocation problem while the CTO evaluates its architecture, the CMO its market position, and the CEO the trade-offs between them.

Humans retain the scarce functions: choosing which futures are worth pursuing, forming relationships, telling the story, accepting accountability, navigating consequential ambiguity, and making judgment calls where being wrong matters.

The objective is not to remove humans from companies. It is to spend human attention where human attention remains scarce.

2. The Compounding Thesis

There are three layers to the argument.

Layer one: search

When the cost of building and operating an experiment falls, the number of futures we can test rises. Early-stage venture outcomes are uncertain. Rather than attempting to eliminate that uncertainty through prediction, DHM seeks to make uncertainty cheaper to resolve.

Build the smallest serious version. Put it in contact with reality. Observe. Kill weak branches. Concentrate on evidence. AI changes the economics of that search.

Layer two: learning

Each company produces operational history: architecture decisions, failed approaches, customer questions, pricing experiments, positioning changes, financial scenarios, product decisions, and technical infrastructure. That history becomes part of the operating corpus available to subsequent companies.

Company N+1 does not begin with a blank prompt. It begins with access to accumulated tooling, workflows, infrastructure, decisions, and lessons from companies 1 through N.

This transfer is already visible in the portfolio. LinuxToaster, OnnAI, and Dokusan share parts of a composable-AI toolchain. Local inference developed in that part of the portfolio now ships in three products. helloMachine and helloMachine Agency approach the same commercial surface through product and services.

The hypothesis is that this inheritance reduces Time to Future with each successive experiment.

Layer three: concentration

Cheap experimentation only creates value if the organization is willing to stop. Most branches should not receive unlimited capital.

DHM's method is to experiment cheaply, expose products to real markets, and concentrate human attention and capital where evidence begins to accumulate. The first phase demonstrated exploration capacity. The next phase is selection.

The important evidence will increasingly be customer adoption, retention, revenue, distribution, proprietary data, trust, and other forms of scarcity that cannot be reproduced merely by generating more software.

3. Portfolio Overview

Company	Category	Description
helloMachine	AI Salesperson	AI salesperson embedded with one script tag. Answers questions, qualifies prospects, books meetings, and can take payment. From $595/mo.
helloMachine Agency	AI Visibility	Audits and monitors how AI systems describe a brand, with AI-oriented content, llms.txt, competitor analysis, and ongoing monitoring.
OnnAI	Iterative Refinement	30-tool toolkit: define what better means, let AI iterate toward it, and retain an auditable record of every change. $2,995 per team.
LinuxToaster	Composable AI	Any model in a Unix pipe, local or cloud. Shared primitives for AI-native workflows. Free through hosted and enterprise deployments.
GameGuide	AI Gaming Companion	Tips and walkthroughs delivered through fan-made game-character mimics. Free and paid consumer tiers.
DCA Partners	Constraint Analysis	AI-augmented consulting built around Dynamic Constraint Analysis: identify, quantify, and relax the constraints limiting system performance.
Dokusan	Conversational Database	Plain-English interaction with private data, implemented in C and designed to run on customer hardware. $2,000–$15,000/mo with hardware.
Aimeru	Relationship Companion	Web assessments and companion, iOS app in beta, and Mac app in beta with local inference for private conversations. Free / $29 mo.

4. What the Portfolio Demonstrates

The current portfolio demonstrates that a very small human organization can build and operate multiple distinct technology products using shared AI-native infrastructure.

It demonstrates technical transfer between companies. Native applications, local inference, developer tools, consumer software, enterprise systems, and commercial products have emerged from the same operating base.

It does not demonstrate that all eight companies should survive. We do not expect them to.

Nor does the current portfolio, by itself, establish venture-scale product-market fit, durable distribution, or portfolio-level financial returns. Those are the tests now underway.

That distinction is central to the model. DHM's thesis is not that AI makes every startup successful. It is that AI makes it substantially cheaper to discover which startups deserve further capital.

5. Where the Moat Can Form

The models are not the moat. DHM uses commercial and open-source models and expects the underlying model frontier to continue changing rapidly.

Operational corpus

History accumulated from operating real companies rather than isolated demonstrations.

Shared infrastructure

Tools, orchestration, deployment systems, local inference, and machine-oriented infrastructure reusable across the portfolio.

Technology transfer

A capability developed for one company can become infrastructure for the next. Local inference is one current example.

Scarce assets

Customer relationships, distribution, proprietary data, trust, domain knowledge, audit history, and positions in emerging standards.

Selection

The willingness and ability to stop funding branches where the evidence does not justify continued attention.

Model capability should commoditize. DHM is built on the assumption that it will.

6. The Economics

Traditional venture building incurs substantial fixed costs before a company has generated enough evidence to justify them: engineering teams, management, operations, market research, design, and go-to-market functions.

AI changes that cost structure. DHM uses AI operators and shared infrastructure to reduce the marginal cost of many of those functions, allowing small human teams to operate multiple experiments.

The economic mechanism is:

lower cost per experiment → more experiments → faster evidence → earlier kills → greater concentration behind the survivors.

The downside of an individual experiment can remain comparatively small while the equity upside of a successful company remains open-ended. That asymmetry is the source of the convexity in the model.

7. Investment Structure

DHM currently offers accredited investors the opportunity to take direct positions in selected portfolio companies. Typical direct positions are $50,000–$200,000.

The companies span AI sales, AI visibility, developer infrastructure, iterative refinement, private data, gaming, consulting, and relationship software. Investors should not assume equal maturity, market evidence, or risk across the portfolio. Individual opportunities should be evaluated on their own terms and through the applicable investment materials.

DHM is also evaluating how the operating stack itself may eventually be commercialized. That is an option, not the core investment thesis today.

8. Principal Risks

Early Stage

The portfolio is young. Building products quickly is not equivalent to finding product-market fit. The principal unanswered question is which companies can convert technical capability into durable customer demand and scarcity.

Selection Risk

Cheap experimentation can become a liability if it produces activity without concentration. The model depends on DHM killing weak branches and allocating disproportionate resources to those supported by evidence.

Founder Dependency

The human founding team remains small. AI reduces the amount of routine cognitive and operational work requiring people, but branch selection, external relationships, accountability, and consequential judgment remain dependent on the founders.

Model and Infrastructure Risk

DHM operates in a rapidly changing ecosystem of commercial and open-source models, hardware, and infrastructure. The architecture is designed to make models replaceable, but capability, economics, licensing, and availability can change.

Regulatory and Liability Risk

AI systems cannot bear legal accountability. Products operating in sensitive or regulated domains require human oversight, appropriate controls, and domain-specific compliance. Regulatory treatment of increasingly autonomous AI systems continues to evolve.

Portfolio Complexity

Shared infrastructure creates leverage, but operating multiple companies also creates coordination and attention risk. The technology tree creates an advantage only if shared components reduce complexity faster than additional companies create it.

Scarcity Risk

Software is increasingly easy to reproduce. A company that fails to accumulate distribution, trust, proprietary data, customer relationships, regulatory position, audit history, or another durable scarce asset may have little defensibility regardless of technical quality.

9. What Investors Should Watch

DHM's model should become more convincing if Time to Future continues to fall as infrastructure accumulates; shared capabilities continue transferring successfully between companies; weak experiments are stopped rather than maintained indefinitely; and a subset of portfolio companies begins showing meaningful customer pull.

The strongest evidence would be surviving companies accumulating scarce assets that competitors cannot reproduce simply by accessing the same models, followed by increasing concentration of capital and human attention behind those companies.

If those things do not happen, the thesis is wrong or incomplete.

A useful venture thesis should expose itself to evidence.

Summary

AI is collapsing the cost of producing many things companies historically paid humans to produce. DHM is built around the possibility that this changes not only software development, but the economics of company formation itself.

We use what AI has commoditized to search a larger technological state space at lower cost. We transfer infrastructure and learning between experiments. We expose those experiments to markets. We stop where evidence fails. Where evidence holds, we seek to own the scarce asset and concentrate capital behind it.

The current portfolio is evidence that the system can produce. The next question is whether it can select and compound.

That is the bet.

Co-investment discussions and partnership enquiries: dhm@dhmventures.com
DHM Ventures · San Mateo, California · Portfolio information current as of September 2026.