Where We Are Investing

Decasonic’s Request for Startups 

-- Abdul Al Ali, Venture Investor, Decasonic

We are at a rapid inflection point in AI acceleration, where AI is moving from delivering outputs to delivering outcomes. In this environment, an interface or application is capable of delivering end to end outcomes, orchestrating a combination of inputs and refined outputs. Where we are right now continues to be human-delegated AI, with individuals and companies providing intents, refining output through feedback and human in the loop, and executing on outputs to derive outcomes. The shift towards AI delivering those outcomes from end to end is what marks the transition point of AI (software and physical) accelerating towards being economic actors.

The struggle in this transition is primarily about ownership of the inputs that deliver outcomes, as highlighted in Satya's recent "The Reverse Information Paradox." Those inputs include codified expertise in the form of files (or skills), workflows, judgement that comes from decades of experience and operations, taste, and the refinement feedback given on outputs. The marginal difference between a median output and a "great output," which has been largely reduced to "taste," is primarily due to domain expertise, and this refinement is what delivers on the transition point of outcome AI. Maintaining ownership of your domain expertise is how companies and individuals will be able to continuously innovate, and avoid the permanent rent trap associated with deployment of frontier, closed-source intelligence.

The environment outlined above assumes the transition point enables two things: (1) ownership of domain expertise, and therefore the availability of user-aligned and user-owned AI, and (2) the ability to price outcome-based AI effectively. Both of these scenarios are being actively addressed. The former is being addressed through a combination of model-orchestration and open-source reliance, where intelligence can be composed and rented while the underlying inputs remain user-owned. The latter is being addressed through initiatives like outcome-based pricing, which align what is paid with what is delivered. We believe this inflection point benefits companies operating at the Web3 x AI intersection, and marks the transition of agents into economic actors.

These value propositions of the Web3 x AI intersection are rapidly manifesting, and are primarily emphasized by (1) agents and machines as economic actors, (2) discoverability associated with software, including means of "production" for agents such as skills, APIs, and workflows, and (3) user-owned and user-aligned AI. This acceleration continues to form the backbone of our investing in the quarter ahead, as Web3 enables the AI economy to scale: as AI moves from generating outputs to delivering outcomes, value accrues to the products that own the user relationship and deliver those outcomes end to end. 

The layer where those value propositions are accelerated is through Web3. Web3 x AI intersection shines through: (1) enabling discoverability of services and software, (2) enabling orchestration of services and software, (3) enabling the settlement of payments for AI agents as they accelerate towards economic actors, (4) enabling agents to deliver outcomes by being "economic actors," and therefore aligning their value with the output being delivered, (5) enabling capital formation associated with agents and AI, where ownership of resources (in the form of revenue share) and ownership of inputs continues to be verified and provided by blockchain rails, (6) distributing expertise, where Web3 allows for the economic coordination of actors involved in the AI economy, enabling it to scale and allowing for the trust and bottleneck of adoption to be addressed in order to delegate and increase the scale of deployment, and (7) enabling the orchestration of outcomes, primarily from the point of providing an intent to receiving an output, which then enhances your ability to receive outcomes. We previously highlighted the sectors we are actively investing in here: link, primarily through our thesis for the upcoming quarter. We aim to highlight the use-cases associated with our sector of deployments in this article. 

Decasonic is a team of investor-operators actively investing in early-stage AI, Web3, and intersection companies. Our team has built 300+ AI agents internally to augment our 3-human investor team members, and we continue to actively share our perspectives through building our AI OS and AI Engines Portal for portfolio companies. We have further mapped hundreds of companies operating at the Web3, AI, and intersection landscape here: link.

Sectors of Interest and Use Cases

Source: Q3 2026 Use Cases of Interest

AI Interfaces and Agents

Adoption of AI agents continued to accelerate in Q2 2026, with the continued advancements made in Ambient, Voice, and Hybrid (hardware and software) AI interfaces. The noticeable trend is the experience associated with AI interfaces, where harnesses are responsible for orchestration and interfaces are responsible for the delivery of outputs. 

Below are some of the use-cases we aim to invest in: 

1. Outcome-Oriented Interface. Regardless of form-factor (ambient, voice, or chat), we are looking to invest in interfaces that are responsible for delivering a complete outcome for users end to end. This includes leveraging the best models, tools, context, and inputs; coordinating them into a finished workflow. Web3 provides the coordination of the inputs required for delivering outcomes. 

2. Multiplayer-Interfaces. AI Interfaces that enables multiplayer-AI experiences and collaboration across domains. This includes placing AI agents in shared spaces, social groups, workspace, or a swarm of coordination agents. Multiplayer enables enhanced experiences, deepened personalization, and compounding retention. 

3. Agentic Wallets. The wallet is the control surface where a human grants an agent bounded, revocable authority to hold and spend. Wallets are increasingly the form factor of identities for AI agents, and the control interface will be represented through Agentic Wallets. 

4. Generative Interfaces. AI is making personalization possible at the level of the interface itself, the same product renders a different layout, flow, or surface for every user and every task. Ownership of the proprietary and usage data behind that personalization is what lets an interface be generated and re-generated on the fly rather than templated once and reused. Web3 provides users with the ability to audit, own, verify, and maintain their own data that enables personalized interfaces. 

Physical AI

Physical AI is actively making towards deployment, with the early innings of mass-scale deployment expected to accelerate this year. 50,000 units of humanoid robots are expected to be shipped this year as the industry continues to accelerate towards mass production. Funding in the sector continues to accelerate, accelerated by both hardware and software funding (including for world and video-model focused labs). Web3 is the coordination layer that lets Physical AI scale its deployment: it works today as resource coordination, connecting the humans, data, sensors and machines a deployment needs, it extends to machine identity that gives each robot or device a verifiable way to be discovered and trusted, and its endpoint is machine-to-machine payments as the machine economy scales, forming the basis of the robotics task economy. 

Some of the use-cases we aim to invest in include: 

1. Machine Coordination and Swarms. The verifiable layer that enables physical AI devices to be discovered and coordinated across manufacturer boundaries. Companies building in this use-case are a necessary precondition for physical-AI focused marketplaces to exist. Fleet operators, insurers, counter-parties, and clients require confirmation on what a device is and what devices are available. This is a use-case that will continue to accrue in value due to the emergence of the “robots-as-a-service,” business model. 

2. Robotic task automation. Robots deployed to complete discrete physical jobs in warehouses, factories, and logistics, sold as robots-as-a-service against measurable ROI on headcount, throughput, and downtime.

3. Machine-to-machine payments. Enabling autonomous transactions between machines. In this scenario, humanoid robots might be able to pay for inference with no human in the loop, with machine to machine payments enabling “machine economic actors,” the extension of agentic software actors. 

4. Sensor Networks and Data Marketplaces. Scaling physical AI depends on real-world sensor and data inputs that no single deployer can cover alone. Web3 is the coordination and capital-formation layer that enables (1) token incentives to allow for a permissionless supply of sensing and data, and (2) the market that prices the usage of the sensor and data in the form of marketplaces. 

AI Applications and Services

In addition to AI Interfaces, AI applications enable the surfacing of outcomes, where the moat is specialized through vertical depth. Services are what an agent hires to do the work, from data and search to tools, evals, and compute. In Q2 2026, we continued to see the “commoditization,” of intelligence driven by (1) rapid advancements from frontier model labs and (2) the adoption of open-source models. Simultaneously, we are seeing the emergence of “picks and shovels,” a layer of the agentic economy being formed. 

Web3 enables AI applications to remain user-owned and user-aligned. Where we are actively looking to back companies include: 

1. Proprietary Domain Expertise Applications. A vertical product that owns one workflow, often differentiated by domains of deployment. This manifests itself in the form of Vertical AI Applications. 

2. Agent Native Services. Services built for AI agents as the end customer, with adoption accelerated by the formation of the agentic economy rails being developed. This includes offering data, search, tools, evals, and compute. 

3. Portable Memory. Gives users ownership of a portable memory layer, their history, judgment, and preferences that follows them across every app and agent, so personalization compounds. 

4. AI Companions and Avatars. Delivering customized, personalized AI experiences through companions and avatars. Web3 enables users to own their experiences and inputs. 

AI Networks, Platforms, and Marketplaces

AI systems need inputs, inference and compute, models, data, tools, and skills, and that need is where the supply layer forms. As agents accelerate toward becoming economic actors in their own right, they buy those inputs from both machines and humans, and networks, platforms, and marketplaces are where that demand gets served. Humans increasingly sell directly into that demand, with domain expertise the input an agent cannot generate on its own but needs to execute on an outcome. Web3 provides the neutral coordination layer that lets these networks compound in adoption while coordinating the inputs required to deliver outcomes.

We are looking to back founders building:

1. Agentic Marketplaces. Autonomy depends on an agent being able to source its own inputs: discover what is available, orchestrate the pieces into a workflow, and pay for what it uses, with minimal human in the loop intervention. Web3 is the neutral coordination layer that makes that possible. 

2. Skills and Distribution Networks. A supply network where capabilities are published, versioned, priced, and called by agents at scale. Web3's capital-formation layer accelerates how fast new skills get built, primarily by coordinating the humans who hold the domain expertise those skills encode.

3. Identity and Reputation Platforms. Provenance an agent can carry with it: a verifiable record of what it is and what it has done, persisted on blockchain ledger. That persistence turns identity and reputation into a portable asset. 

4. RL Marketplaces. Marketplaces where human domain experts sell RLHF and evaluation services directly to the AI agents and models that need judgment in domains that require judgement. Web3 scales how many experts a marketplace can recruit and pay across borders.

AI x Fintech

Capital and settlement is increasingly becoming the constraining factor for the acceleration of agentic economies. The agentic finance layer is where value accrual will be pooled, settled, and formed.

We are looking to back founders building:

1. Agentic Treasuries. Treasury management enables solutions to be jointly by humans and AI: the agent handles the continuous work of allocating idle capital, rebalancing positions, and executing routine transfers, while a human sets the policy and holds override authority. 

2. Self-Sustaining Agents. An agent pre-funded with yield-bearing capital that earns enough to cover its own compute, task execution, and transactions, so it keeps operating without a human topping it up along the way.

3. Agent to Agent Commerce. The exchange layer where one agent buys from or sells to another, with agentic escrow being amongst the core primitives needed to accelerate adoption by enabling both trust and scale. 

4. Compute Finance. AI deployment is accelerating the capital expenditure hyperscalers commit to the buildout, roughly $750B in capex this year and a projected $1.1T next. Tokenized computer markets let a buyer lock in that computer at a pre-defined price through a forward contract, hedging the spend.

Consumer AI

Consumer AI is about delivering personalized experiences, where the differentiator is the context a product accumulates about a user and the relationship graph that context compounds into with usage. Giant, a portfolio company, builds personalized experiences for children, turning each child into the star of their own story generated around them, and it is the personalization thesis in its most defensible form: the more a child creates, the more the product knows, and the harder it is to leave. Web3 turns that context and memory into user-owned, portable assets that follow the person across products, keeping the experience aligned to the user rather than the platform.

Use-cases we are looking to invest in include:

1. Multiplayer AI Experiences. An AI that lives inside a shared space, a social group or a workspace, where shared memory and co-owned agents allow for enhanced experiences. 

2. Personalization Loop. A consumer product where the more a user creates, the more it knows and the harder it is to leave. This includes across both productivity and entertainment use-cases. 

3. Tokenized Attention Markets. Marketplaces that turn the attention and engagement a user creates into an ownable, tradable asset. Others finding value in the asset can therefore participate and own it. 

4. Social AI. Enabling users to form relationships with their agents, with the exchanges building shared memory, context, and history that builds relationship graphs. Web3 enables that agent, and everything it has learned about the user travels across interfaces and applications. 

Conclusion 

We are an early-stage investor-operator team, and we invest as operators who have built the agents and applications ourselves. We are actively investing across these six sectors, and across the use-cases highlighted in this article. If you are building the frontier of Web3 x AI and you want a value-add, investor-operator partner, reach out to us at Decasonic - we are actively deploying capital. 

The content of these blog posts is strictly for informational and educational purposes and is not intended as investment advice, or as a recommendation or solicitation to buy or sell any asset. Nothing herein should be considered legal or tax advice. You should consult your own professional advisor before making any financial decision. Decasonic makes no warranties regarding the accuracy, completeness, or reliability of the content in these blog posts. The opinions expressed are those of the authors and do not necessarily reflect the views of Decasonic. Decasonic disclaims liability for any errors or omissions in these blog posts and for any actions taken based on the information provided.