Content List
DeHAIS or DeVAIS – this is the question
Nick Smohorzhevskyi
14.02.2025
12 Min Read
Try googling WTF are DeHAIS and DeVAIS first. You’ve probably found nothing. Luckily, we are head-deep in product research in Solus, so let’s implement these new narratives that will join us in the rally soon.
So, here is the thing: Both Vertical and Horizontal Swarms were under the radar from the middle of 2024 if not earlier for the peeps sitting neck in AI developments. Still not perfect but visible, this narrative found its founders in October. Now we start to see the ways to execute that in business logic, even though it is not so perfect, but with the idea to collaborate on further developments.
DeHAIS
DeHAIS is a decentralized system where autonomous AI agents work side by side, handling a variety of tasks without a strict hierarchy.
These agents are designed for generalized task handling and share responsibilities across a distributed network. The key feature of DeHAIS is its ability to dynamically distribute tasks among agents based on availability, capability, and environmental inputs, making it suitable for broad, multi-purpose applications.
DeVAIS
DeVAIS operates as a domain-specific swarm framework, focusing on vertical integration of specialized AI agents. These agents are tailored for precise tasks within a specific domain, such as finance, healthcare, or cybersecurity. The system leverages a hierarchical structure to optimize performance in specialized contexts, enabling enhanced precision and efficiency.
VC Insight:
Your next 100x opportunity is waiting for you in this article — we even went ahead and gave it its own headline!
So, story time:
Imagine you’re walking into a bustling city, where every aspect of life —business, healthcare, transportation, and security — runs smoothly, powered by two kinds of advanced AI systems: DeHAIS and DeVAIS. These systems aren’t just theories, they’re frameworks transforming how work gets done, each designed for a specific type of task.
DeHAIS: The Jack-of-All-Trades
Picture a network of autonomous drones and bots spread across your city. You may or may not have them already, and this is only the beginning. These agents don’t have a centralized command centre. Instead, they communicate via a trusted communication network to figure out who does what, based on availability, skills, and what’s happening around them.
For instance:
- Logistics and Delivery:
We had a discussion with a logistics company – and even with their focus on digitizing the logistics industry, they can often be stuck with a last-yard delivery track due to narrow streets and corridors and unpredictable cars and foot traffic. But when you imagine a fleet of small delivery drones and bikes – everything gets way smoother. These are DeHAIS agents.
The moment an online order drops, these agents get to work, figuring out:
- Which drone is the closest to the pickup point;
- Which route avoids the latest traffic jams and crowds;
- How to group orders that go to the same area to save time.
If one bike gets a flat tyre, the system automatically assigns another nearby agent to take over without disrupting the flow.
- Smart Manufacturing:
While we already have automated factory lines, these machines and robots are not part of an interconnected network. One day, a welding robot breaks down in the middle of an important production run.
Instead of stopping the entire line, DeHAIS agents step in:
- A nearby assembly robot temporarily takes over the welding task, using its adaptable tools;
- Other agents reassign themselves to balance the workload, ensuring that the rest of the production line keeps moving;
- Meanwhile, a maintenance agent diagnoses and repairs the broken robot, restoring it without human intervention.
This system ensures the factory operates without interruptions, reducing downtime and maximizing efficiency.
Before our research at Solus, we engaged with logistics providers and manufacturing innovators regarding what can be the missing layer in automation. We help create ideas around AI Swarm potential with future integrations into real-world applications. And while it can be a slow process, we are accelerating.
VC Insight:
If you close here – you lose an opportunity to fulfil the KPIs of your Fund in less than a year. 3X in 5 years is easy.
10 or 15 years ago both concepts were quite futuristic. Now we have delivery drones, self-driven taxis and automated manufacturing lines. While this is just the beginning, we are slowly entering the world of automated operations, where the level of success for task execution gets higher with each update.
DeFi moved financial transactions to smart contracts. DeHAIS and DeVAIS are doing the same for decision-making and automation across industries. Decentralized storage, computation and smart contracts are laying the groundwork for fully autonomous AI swarms.
And while logistics and any real-world business may be just boring you can optimize your workflow in Web3 on both sides.
What is the next phase? AI-powered agents execute on-chain logic without human inefficiencies, and investors funding this now are front-running mass adoption.
But where to seek it? DeFAI?
And while on the business end, it may be pretty obvious how DeFi platforms using DeHAIS to manage liquidity pools and mitigate risks.
Where DeHAIS agents monitor pool activity in real time, reallocating funds between liquidity pools to balance supply and demand. If a flash loan attack is detected, DeHAIS agents immediately stop abnormal transactions and trigger protective mechanisms via smart contracts.
And say even more – this way you can also dynamically adjust staking rewards, incentivizing liquidity providers based on market trends.
On degen plays it can become crazy as with an example of Flash Loan Farming Swarms.
Imagine automated DeHAIS agents optimizing flash loan strategies to exploit arbitrage across DeFi protocols.
VC Insight:
We will have such solutions soon, stay tuned. Arbitrageurs are starting to build their infra, no jokes.
They are quick and continuously monitoring DEXs for price discrepancies between tokens. When a profitable arbitrage opportunity arises, an agent executes a flash loan to buy low and sell high across multiple DEXs within a single block. And while experienced users can say it won’t work, you as a reader can be the first one to check and implement it. The one with better math and software wins – so it is really an opportunity.
But let’s talk about the other use cases where you need to sell a big portion of tokens – DeHAIS agents can coordinate collective ladder selling to make it smooth. So, anything can just become better and faster when you need multiple actions done via the same algorithm. So imagination is your next step before implementation.
The secret to DeHAIS’s flexibility is its distributed decision-making
Each agent has the intelligence to assess its own capacity and collaborate with others to complete tasks efficiently. This decentralized approach eliminates bottlenecks and adapts quickly to changing environments.
DeVAIS: The Specialists
Now imagine another part of the city (oh well, here we go again, some random futuristic city), where precision and expertise are the key. DeVAIS agents aren’t generalists like DeHAIS — they’re highly specialized in specific industries. These agents work in hierarchies, with some focused on big-picture strategies and others handling detailed tasks.
Let’s start with some examples to understand the concept:
- Healthcare: In hospitals, DeVAIS systems manage patient care. High-level agents analyze patient data to predict treatment outcomes, while specialized bots administer medication, monitor vitals, and alert doctors if anything seems off;
- Finance: In financial institutions, DeVAIS agents oversee trading, fraud detection, and portfolio management. One layer monitors market trends, while others execute trades or flag suspicious activity in real time;
- Cybersecurity: In the city’s data centres, DeVAIS agents continuously scan networks for vulnerabilities. High-level agents simulate potential attack scenarios, while others apply fixes or isolate compromised systems.
The easiest way to implement a closer way to Web3 is:
Different DeFAI and Market Analytics
Market analytics in Web3 is ideal for enhancement with DeVAIS. Specialized operational agents can focus on areas like Lending and Borrowing rates, and floor prices for the NFTs or just provide you with sentiment analysis and gathering data, analyzing trends & summarizing insights.
For example, a DeFi-oriented operational agent tracks liquidity and protocol updates, while an NFT-oriented operational agent monitors sales volumes and trending collections. The hierarchically higher Agent refines this input, filters out the noise, and delivers actionable reports that matter.
On DeFAI – the problem of rebalancing and refinancing can be properly automated. Markets are 24/7, yields shift constantly, and risks change by the minute. DeVAIS automation that dynamically optimizes portfolios, reallocates assets and manages to refinance with zero human input can be a key factor here. Imagine a system that detects underperforming assets, shifts funds to higher-yield pools, and secures better lending rates — all in real time. No sleep. No human error. Just relentless optimization.
At Solus, we work with DeFAI projects, like GraphDEX to build around on-chain automation and simplification. The emergence of AI-driven simplified finance isn’t a question of if, but how seamlessly it integrates, so…
VC Insight:
Start implementing or just invest in Trading Terminals that have plans to integrate that.
So if you are a Web3 Hedge Fund it can be the best timing to take a look at the AI space to be among the first ones to implement new solutions once they are there.
To say it shortly, DeVAIS thrives in environments where domain expertise is critical. Its hierarchical structure ensures that specialized tasks are completed efficiently while strategic decisions remain aligned with broader goals.
Here we can move to a comparison.
Horizontal vs Vertical AI Swarms vs The World
- Horizontal AI is your everyday AI — broad systems which can handle a range of tasks but aren’t too specialized;
- Vertical AI is tailor-made for a specific industry. Such AI covers one narrow complex of tasks, or only 1 complex task.
Horizontal Swarms = decentralized, peer-to-peer agent networks. No hierarchies, no bosses. Every agent just figures it out in real-time. Tasks get distributed dynamically based on local conditions or decentralized signals.
Horizontal swarms can be used for tasks requiring general-purpose collaboration, such as network optimization or can be used in cybersecurity:
- Broad systems which can handle a range of tasks but aren’t too specialized;
- Cybersecurity systems, where agents collaborate to detect and mitigate threats in real time.
Vertical Swarms = structured, hierarchical agent collectives. Specialized agents do their one job perfectly, then pass the results up the chain for bigger-picture decision-making.
Vertical Swarms are designed for industry-specific needs, such as healthcare diagnostics or financial risk assessment:
- Financial modelling, where specialized agents analyze market trends, execute trades, analyze tweets from that one influencer who can move markets with an emoji, and assess risks in a coordinated hierarchy;
- Or to be more real - rescue operations in post-disaster scenarios, where drone swarms can coordinate search-and-rescue efforts: one has a thermal camera, one carries medical supplies, while another blasts “Don’t worry, be happy” to keep morale up.
Communication Protocols and TEEs
This publication was not planned to be about Protocols or Execution Environments, but without them, it would be unreasonable to talk about swarms or anything automated but semi-confidential. Most have seen that Black Mirror series, especially the one about hacked mecha-bees, that started to kill people and you definitely don’t want to die from a robot, who delivers pizza. This is why we need Communication Protocols and TEEs. Because Agents need to communicate, and you shouldn’t be able to listen and interrupt.
So, how do AI SWARMS talk without getting REKT?
There are 2 ways:
- Secure Peer-to-Peer Protocols — Agents whisper sweet nothings to each other using Libp2p and other decentralized networking stacks. No middlemen, no bottlenecks. Just straight-up encrypted communication.
- Low-latency — Because waiting is for noobs
- No single points of failure — Try DDoSing this
- Everything is Encrypted — Even if someone intercepts, we all know what they’ll see
- Blockchain-Integrated Comms — Agents leave on-chain receipts for every move they make. No cap.
- Task histories, status updates, and resource allocations = on-chain. No room for excuses
- Smart contracts = auto-verification. No human mods needed
- Zero-trust, 100% tamper-proof. Basically, AI agents with ledgers
And what even is a TEE? Why do we need it?
So, to make it simple – TEEs make sure no one leaks alpha, so idk how you are reading that, but anyway.
TEE 101:
- Runs computations in a hardware-isolated vault — like a VIP section for sensitive AI work;
- Stops nosy hosts from peeking at secret sauce — your AI agent’s alpha remains hidden;
- No backdoors = no funny business — even the sysadmin is not invited.
And actually thanks to TEEs it will be possible to work with sensitive data, like in the healthcare industry, or at least people won’t know you like pineapple pizza.
VC Insight:
Hey, it is below, check it.
Morphic
Is someone building that? Yeah, guys from Morphic is basically a 𝗧𝗿𝘂𝘀𝘁𝗹𝗲𝘀𝘀 𝗮𝗽𝗽𝗰𝗵𝗮𝗶𝗻 𝘀𝗲𝗿𝘃𝗶𝗰𝗲 (𝗧𝗔𝗦), an enhanced TEE AVS that first-ever presents not yet presentable 𝘀𝗲𝗰𝘂𝗿𝗶𝘁𝘆 (L1-enforced at-least-one honesty assumption), SaaS-like 𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆 (100ms verifiability and finalization), and 𝗽𝗿𝗼𝗴𝗿𝗮𝗺𝗺𝗮𝗯𝗶𝗹𝗶𝘁𝘆 (Docker-based).
So, now you need something new, that you didn’t know existed, and you didn’t know you even needed it. Ping me on TG for an intro or just DM to invest early.
Map of the DeHAIS/DeVAIS Ecosystem
Just check the projects below and see if you like something to try… Just try it, mate.
Theoriq — a modular and composable framework that allows the creation of AI agents & Swarms focused on the DeFi sector, creating DeFAI. Next-gen project advancements include: complex data processing, workflow automation, agentic DAOs, and trading swarms.
Intermittent Fasting Science — combines DeSci and AI agents (“DeSAI”) in a vertical agentic system to facilitate intermittent fasting and healthy eating. IFSCI’s Multi-Agent Control (MAC) system enables collaborative AI agents to address complex challenges: users can upload meal photos via social media or the in-house app for automated nutritional analysis and track fasting periods.
SwarmZero — facilitates the creation of dApps by orchestrating swarms of AI agents. These agents collaborate to automate the development process, reducing the need for manual coding. Requires API keys from OpenAI, Anthropic, and Mistral.
FXN World — protocol for agent collaboration, functions utilizing the “Resource agents” layer that provide other agents with access to data, APIs, computation, or the results of actions performed by other agents.
Naoris Protocol — enhances cybersecurity in both Web2 and Web3 environments, utilizing the AI agent Swarm that monitors network activities, detects anomalies, and responds to threats in real time.
ZAI — multi-agent orchestration framework designed for autonomous AI-driven onboarding, trading analytics, and Web3 adoption, leveraging on-chain and off-chain data integration. Its self-evolving agent network dynamically optimizes performance through distributed collaboration, real-time data processing, and reinforcement learning, with user AI agents acting as adaptive nodes within the ELIZA framework.
ElizaOS — AI agent operating system designed to facilitate the creation and orchestration of AI agents & Swarms within the Web3 ecosystem. Its modular plugin architecture allows for the integration of various services, including web browsing, image analysis, and blockchain interactions, enabling agents to perform complex tasks collaboratively.
Swarms World — multi-agent AI and LLM orchestration framework with modular architecture designed to automate real-world activities at scale within the Web3 ecosystem. It enables the creation, management, and deployment of AI agents, models, prompts, tools, and swarms, facilitating collaborative interactions among agents to execute diverse tasks autonomously.
Questflow — orchestration layer for the multi-agent economy and development of intent-based AI agents designed to execute user-defined tasks within Web3 ecosystems. Agents operate autonomously, collaborating to perform functions such as data analysis, content creation, market monitoring, and community engagement.
Swarmlabs — Vertical AI Swarm that covers content creation for creators and social networks.
SwarmNode.AI — serverless execution platform optimized for AI agents, providing dynamic orchestration, auto-scaling, and dependency management for Python-based workloads. It enables developers to deploy AI agents without infrastructure overhead, offering features like inter-agent chaining, REST API and SDK-based orchestration.
FixieAI — platform for creation of AI Agents that can be connected to various business apps and data sources or each other, creating a swarm. Fixie encourages developers to build smaller, specialized agents that collaborate to solve bigger problems.
Langchain — an open-source framework that makes it easier to build applications composed of multiple AI agents, each performing discrete tasks (e.g., data retrieval, summarization, decision-making), and passing outputs to one another. Reinforced by LangSmith LLM framework and LangGraph platform for control of agents.
Harvey & EvenUp — AI solutions for the legal industry that consist of multiple AI agents focused on specific legal tasks and issues creating swarms. Platforms utilize natural language processing to assist with legal research, document drafting, and case analysis, streamlining workflows for legal professionals.
Adept AI — building an AI assistant capable of performing actions across multiple software tools, websites, and APIs. Their model (dubbed “ACT-1”) focuses on orchestrating tasks autonomously — e.g., booking meetings, extracting and analyzing data, or making edits in productivity apps — by coordinating smaller specialized functionalities.
OpenAI — OpenAI’s Swarm modular framework is an experimental, lightweight platform designed to facilitate the creation and orchestration of AI agents capable of collaborative task execution. It provides developers with tools to construct multi-agent systems where agents can communicate and delegate tasks among themselves, thereby streamlining workflows across various applications.
Conclusion: The AI Swarm Era Has Begun – Now What?
The rise of DeHAIS and DeVAIS is not just another AI trend (or it is), but it is the general evolution of automation, decision-making, and decentralized intelligence. AI Swarms are not theoretical anymore. They are being built, optimized, and integrated into supply chains, DeFi strategies, cybersecurity networks, and even autonomous trading. The question is no longer if AI Swarms will take over key Web3 infrastructures – it’s how fast.
For Users, the opportunity is clear: leverage AI agents to automate on-chain interactions, optimize yield strategies, and enhance data-driven decision-making. Whether it’s DeFAI-driven arbitrage, AI-enhanced governance, or automated risk assessment, Swarm Intelligence enables Web3-native self-optimization with minimal human interference. Early adopters will outperform manual DeFi participants and execute at a speed impossible for individuals.
For Businesses, integrating DeHAIS and DeVAIS is not a futuristic bet — it’s a competitive necessity. AI Swarms reduce inefficiencies, enhance security, optimize liquidity management, and eliminate operational bottlenecks. Enterprises need to adopt agent-based systems early to ensure their platforms remain scalable, exploit automated decision-making, and leverage tokenized AI participation models for self-sustaining ecosystems.
For VCs, this is a once-in-a-cycle asymmetric bet. Investing in modular, decentralized AI agents before mass adoption will return exponential network effects as the infrastructure matures. We already have the first iterations of what will soon be a dominant force in automation. The question isn’t should AI Swarms be funded – it’s which projects are best positioned to dominate. And has a real tech, yeah, don’t forget about it.
Whether you’re a founder developing AI-driven automation, a Web3 company looking for market integration, or an investor seeking high-growth AI opportunities – feel free to reach out.
Either you build, integrate, or invest – or you watch from the sidelines. Choose wisely. DYOR. Solus.
AI Swarms
TEE
Decentralization
VC Investments
Automation
Share this post
