Op-Ed · AI Infrastructure
The next phase of AI adoption will not be defined by better models. It will be defined by the workspace, controls, and infrastructure layer that turn an agent into something a business can actually run.
Cedral Advisory · May 2026 · 11 min read
Op-Ed
AI
Infrastructure
Most AI products still treat the agent as a chat window.
A user opens a session, asks a question, gets an answer, and leaves. That experience can be useful, but it is not the same thing as having a durable AI teammate that understands a business, holds context, follows operating rules, and can be trusted inside a real workflow.
The next phase of AI adoption will not be defined by better models alone. It will be defined by the infrastructure and product layers that make persistent agents useful, manageable, and safe for businesses.
That is the opportunity Cedral is building toward.
A model API can answer a prompt. A business agent needs much more around it.
It needs an identity. It needs role definition. It needs company-specific context, user preferences, and a persistent operating profile. It needs permissions, access control, billing, usage limits, workspace diagnostics, memory review, deletion controls, and a recovery path when the runtime fails. Then, on top of all of that, it needs a user experience that a normal business can actually understand.
Without that layer, an “agent” is just a stateless chatbot with better branding.
“For agents to become real business infrastructure, the agent needs a home. A workspace, a control plane, a profile, a memory model, and an operational wrapper.”
That is the layer most of the market is skipping.
There is another reason this matters right now: the AI stack is changing too quickly for static infrastructure assumptions.
Models are improving. Agent frameworks are evolving. Security expectations are rising. Customers are beginning to ask harder questions: where their data goes, how persistent systems behave, and what happens when the underlying runtime needs to change.
AI infrastructure cannot be rigid. Businesses need a layer where agent runtimes can be deployed, resized, migrated, upgraded, or shut down quickly as the stack underneath evolves.
This is where Akash becomes interesting.
The practical Akash thesis is not simply that decentralized compute is cheaper, though cost can matter. It is not that every business suddenly wants “decentralized AI” as a slogan. The stronger thesis is adaptability.
Persistent agents benefit from compute portability, dedicated runtimes, lower-cost always-on infrastructure, and reduced dependence on any single hyperscaler’s primitives. For AI-native businesses, protocol teams, research firms, and any company that cares about operational sovereignty, that flexibility is going to matter more over time, not less.
Persistent agents are not just inference calls
A dedicated agent needs its own runtime, filesystem, tools, queue, health checks, and persistent environment. That makes decentralized compute meaningfully more relevant than it is for a stateless chat completion request.
The right framing is careful. This is not “fully private by default.” It is not “on-chain AI agents.” It is not “no cloud middleman.”
The better language is more grounded: dedicated runtime, portable infrastructure, operational sovereignty, adaptable deployment layer, business-grade agent workspace, decentralized compute-backed deployment.
Akash provides the runtime substrate. Cedral provides the business-facing layer.
This is no longer theory. The Dedicated Agent beta is live in production.
The current architecture includes an Akash-hosted Hermes agent deployment, a Cedral server-side bridge between the portal and the agent, authenticated profile and workspace sync, user-controlled operating profile context, workspace status diagnostics in the Agent Console, metadata-only sync visibility that never exposes private workspace contents, chat-clearing semantics that preserve profile and workspace context, and production health checks behind auth-gated endpoints.
This is not the final architecture. Autonomous long-term memory is not live yet. Per-user deployment and automated provisioning still need hardening. The product is appropriately labeled beta.
But the shape is now visible.
“A business user does not need to understand Akash leases, SDL files, provider selection, bridge tokens, or agent containers. They interact with a branded AI workspace. Behind the scenes, the agent runtime lives on decentralized infrastructure. That abstraction is the product.”
The business layer is everything that turns raw agent infrastructure into something a company can actually buy and use: onboarding, authentication, company workspaces, billing, profile and context management, memory controls, admin permissions, diagnostics, support workflows, security boundaries, model and runtime routing, and deployment lifecycle management.
The model is important. The runtime is important. But the business layer determines whether the agent becomes part of daily operations, or whether it stays a curiosity.
This is where a serious portion of long-term value will accrue.
Cedral is not positioning Akash-hosted agents as magically private by default. Today, the value is dedicated runtime, portability, operational control, and an adaptable deployment model.
But trusted execution environments (TEEs) are a meaningful future unlock. A Cedral Dedicated Agent running inside a TEE-backed Akash environment could eventually combine a business-facing AI workspace with dedicated, portable, confidential compute. That would give businesses stronger execution guarantees while preserving the product layer they actually need: onboarding, workspace controls, memory management, diagnostics, and recovery.
That is not a claim about what is fully solved today. It is a roadmap for where business AI infrastructure can go.
The AI infrastructure market is moving fast toward agentic systems, but most products still collapse into one of two categories: generic chat interfaces wrapped around a model API, or technical agent frameworks that normal businesses cannot operate.
There is significant room between those extremes.
Many businesses will want persistent AI agents, but they will not want to operate agent infrastructure themselves. They will want the outcome: a durable, configurable AI workspace that understands their company, respects their controls, and evolves as the underlying stack changes.
That is especially interesting for Akash, because it gives decentralized compute a real business-facing path into usage. Not as an abstract infrastructure story. As a product story.
The near-term roadmap is clear: continue testing the Dedicated Agent beta in production, harden provisioning and approval workflows, design autonomous memory with review, edit, and delete controls, clarify per-user and per-company Akash deployment topology, build recovery and backup paths, track the TEE and confidential compute roadmap, and package the architecture into a concise external demo.
The long-term thesis is simple.
“Every company will eventually have persistent AI agents. The winners will not just provide models. They will provide the workspace, the controls, the trust layer, the infrastructure adaptability, and the operational reliability around those agents.”
Cedral is building that business layer.
Akash may be one of the most interesting places to run the agent runtime underneath it.
The views expressed in this op-ed are those of Cedral Advisory and are provided for informational and educational purposes only. Nothing in this piece constitutes financial or investment advice. Always conduct your own research before making any investment decisions. Note: AI was used in the sourcing of this information.
Topics
AI Agents
Akash
Decentralized Compute
Cedral AI
TEE
Op-Ed
Op-Ed · Web3 Gaming
Web3 gaming receives more hate than it deserves — and the people throwing shade are missing one of the most obvious use cases in crypto right in front of their faces.
Web3 gaming receives a lot of hate and praise from members in the crypto space. Some of this treatment is understandable — most of it is horribly unjustified. Like most sectors in crypto, bad actors and profit-chasing schemers were quick to pollute what should have otherwise been a beacon of excitement and hope for gamers globally.
Endless hours are poured into video games of all genres with one congruent theme: money put in is money lost forever.
“Those toys you played with as kids — dolls to some, action figures to others — luckily hold some, if not gain, value as you grow older. The same cannot be said for the poor fools who spent hours, days, weeks, or even months of cumulative time playing video games.”
Those hours were lost to the black hole of an archaic, centralized system. And it didn’t have to be that way.
Currently, the gaming industry is worth more than the global box office and music industry combined. That’s not a typo. Gaming — and I use that term broadly, because there are quite a few different slices that make up this pie — is an objectively massive market. Mobile games alone make up roughly 50% of it, with console and PC splitting the remainder.
With no sign of slowing down, the magnitude of major titles is increasing. Games like GTA VI are leaving consumers salivating. The average GTA V player logged somewhere between 55 and 70 hours — roughly 2 to 3 days of their life. We all know someone with a few weeks on the clock. When your business revolves around a consumer base that is rapidly growing and staying for longer, you are in a very good market.
This point has been recognized by the people running these companies. Strauss Zelnick, CEO of Take-Two Interactive, has floated the idea of a pay-per-hour cost structure rather than an upfront game purchase. Respect to the thinking. I have a better idea, Strauss.
Let’s embrace gaming’s natural destiny: blockchain integration. Not as a gimmick. Not as a speculative layer tacked onto a mediocre game. The right model is what Off The Grid has demonstrated — use a traditional engine for the game itself, and blockchain exclusively for the in-game marketplace and item collection. Game first. Blockchain second.
This solves two problems simultaneously. Your friend who spent weeks grinding the game can actually benefit monetarily from his time. And companies like Take-Two can still make money — more money, arguably — off marketplace transaction fees. They already rake in billions from microtransactions. They still will. People will still be using fiat to bridge into whatever the in-game currency is, and developers can take a healthy cut of every peer-to-peer transaction.
The math works for everyone. A man who is starving will not care if you keep half the 12-ounce steak you are offering him. Gamers won’t care either — because any system that gives them real ownership of their digital assets is so fundamentally better than what exists today that the fees are irrelevant.
While some are still missing the forest for the trees, there is a gigantic use case for Web3 gaming and NFTs that is right in front of our faces. Everyday gamers — the guy grinding Warzone, MyPlayer in 2K, Fortnite, PUBG, Rocket League, whatever your game of choice is — are craving a system like this. They just don’t know it yet.
“Imagine a game that runs high-stakes tournaments where your favorite streamer is competing with real money on the line — for skins, gear, NFTs with actual market value. More viewers. More donations. More engagement. The whole gaming machine gets bigger.”
The seeds of this future are already planted. Cedral Advisory will be publishing a full research report on the state of Web3 gaming, with an in-depth look at Gunzilla Games and Off The Grid, the role of the industry giants, and what the NFT ownership thesis really means for the next decade of digital entertainment. Stay tuned.