TokenRing AI

Tag: Cognition AI

  • The Rise of the Agentic IDE: How Cursor and Windsurf Are Automating the Art of Software Engineering

    The Rise of the Agentic IDE: How Cursor and Windsurf Are Automating the Art of Software Engineering

    As we move into early 2026, the software development landscape has reached a historic inflection point. The era of the "Copilot"—AI that acts as a sophisticated version of autocomplete—is rapidly being eclipsed by the era of the "Agentic IDE." Leading this charge are Cursor, developed by Anysphere, and Windsurf, a platform recently acquired and supercharged by Cognition AI. These tools are no longer just suggesting snippets of code; they are functioning as autonomous engineering partners capable of managing entire repositories, refactoring complex architectures, and building production-ready features from simple natural language descriptions.

    This shift represents a fundamental change in the "unit of work" for developers. Instead of writing and debugging individual lines of code, engineers are increasingly acting as architects and product managers, orchestrating AI agents that handle the heavy lifting of implementation. For the tech industry, the implications are profound: development cycles that once took months are being compressed into days, and a new generation of "vibe coders" is emerging—individuals who build sophisticated software by focusing on intent and high-level design rather than syntax.

    Technical Orchestration: Shadow Workspaces and Agentic Loops

    The leap from traditional AI coding assistants to tools like Cursor and Windsurf lies in their transition from reactive text generation to proactive execution loops. Cursor’s breakthrough technology, the Shadow Workspace, has become the gold standard for AI-led development. This feature allows the IDE to spin up a hidden, parallel version of the project in the background where the AI can test its own code. Before a user ever sees a proposed change, Cursor runs Language Servers (LSPs), linters, and even unit tests within this shadow environment. If the code breaks the build or introduces a syntax error, the agent detects the failure and self-corrects in a recursive loop, ensuring that only functional, verified code is presented to the human developer.

    Windsurf, now part of the Cognition AI ecosystem, has introduced its own revolutionary architecture known as the Cascade Engine. Unlike standard Large Language Model (LLM) implementations that treat code as static text, Cascade utilizes a graph-based reasoning system to map out the entire codebase's logic and dependencies. This allows Windsurf to maintain "Flow"—a state of persistent context where the AI understands not just the current file, but the architectural intent of the entire project. In late 2025, Windsurf introduced "Memories," a feature that allows the agent to remember specific project-specific rules, such as custom styling guides or legacy technical debt constraints, across different sessions.

    These agentic IDEs differ from previous iterations primarily in their degree of autonomy. While early versions of Microsoft (NASDAQ: MSFT) GitHub Copilot were limited to single-file suggestions, modern agents can edit dozens of files simultaneously to implement a single feature. They can execute terminal commands, install new dependencies, and even launch browser instances to visually verify frontend changes. This multi-step planning—often referred to as an "agentic loop"—enables the AI to reason through complex problems, such as migrating a database schema or implementing an end-to-end authentication flow, with minimal human intervention.

    The Market Battle for the Developer's Desktop

    The success of these AI-first IDEs has sparked a massive realignment in the tech industry. Anysphere, the startup behind Cursor, reached a staggering $29.3 billion valuation in late 2025, reflecting its position as the premier tool for the "AI Engineer" movement. With over 2.1 million users and a reported $1 billion in annualized recurring revenue (ARR), Cursor has successfully challenged the dominance of established players. Major tech giants have taken notice; NVIDIA (NASDAQ: NVDA) has reportedly moved over 40,000 engineers onto Cursor-based workflows to accelerate their internal tooling development.

    The competitive pressure has forced traditional leaders to pivot. Microsoft’s GitHub Copilot has responded by moving away from its exclusive reliance on OpenAI and now allows users to toggle between multiple state-of-the-art models, including Alphabet (NASDAQ: GOOGL) Gemini 3 Pro and Claude 4.5. However, many developers argue that being "bolted on" to existing editors like VS Code limits these tools compared to AI-native environments like Cursor or Windsurf, which are rebuilt from the ground up to support agentic interactions.

    Meanwhile, the acquisition of Windsurf by Cognition AI has positioned it as the "enterprise-first" choice. By achieving FedRAMP High and HIPAA compliance, Windsurf has made significant inroads into regulated industries like finance and healthcare. Companies like Uber (NYSE: UBER) and Coinbase (NASDAQ: COIN) have begun piloting agentic workflows to handle the maintenance of massive legacy codebases, leveraging the AI’s ability to "reason" through millions of lines of code to identify security vulnerabilities and performance bottlenecks that human reviewers might miss.

    The Significance of "Vibe Coding" and the Quality Dilemma

    The broader impact of these tools is the democratization of software creation, a trend often called "vibe coding." This refers to a style of development where the user describes the "vibe" or functional goal of an application, and the AI handles the technical execution. This has lowered the barrier to entry for founders and product managers, enabling them to build functional prototypes and even full-scale applications without deep expertise in specific programming languages. While this has led to a 50% to 200% increase in productivity for greenfield projects, it has also sparked concerns within the computer science community.

    Analysts at firms like Gartner have warned about the risk of "architecture drift." Because agentic IDEs often build features incrementally based on immediate prompts, there is a risk that the long-term structural integrity of a software system could degrade. Unlike human architects who plan for scalability and maintainability years in advance, AI agents may prioritize immediate functionality, leading to a new form of "AI-generated technical debt." There are also concerns about the "seniority gap," where junior developers may become overly reliant on agents, potentially hindering their ability to understand the underlying principles of the code they are "managing."

    Despite these concerns, the transition to agentic coding is viewed by many as the most significant milestone in software engineering since the move from assembly language to high-level programming. It represents a shift in human labor from "how to build" to "what to build." In this new landscape, the value of a developer is increasingly measured by their ability to define system requirements, audit AI-generated logic, and ensure that the software aligns with complex business objectives.

    Future Horizons: Natural Language as Source Code

    Looking ahead to late 2026 and 2027, experts predict that the line between "code" and "description" will continue to blur. We are approaching a point where natural language may become the primary source code for many applications. Future updates to Cursor and Windsurf are expected to include even deeper integrations with DevOps pipelines, allowing AI agents to not only write code but also manage deployment, monitor real-time production errors, and automatically roll out patches without human triggers.

    The next major challenge will be the "Context Wall." As codebases grow into the millions of lines, even the most advanced agents can struggle with total system comprehension. Researchers are currently working on "Long-Context RAG" (Retrieval-Augmented Generation) and specialized "Code-LLMs" that can hold an entire enterprise's documentation and history in active memory. If successful, these developments could lead to "Self-Healing Software," where the IDE monitors the application in production and proactively fixes bugs before they are even reported by users.

    Conclusion: A New Chapter in Human-AI Collaboration

    The rise of Cursor and Windsurf marks the end of the AI-as-a-tool era and the beginning of the AI-as-a-teammate era. These platforms have proven that with the right orchestration—using shadow workspaces, graph-based reasoning, and agentic loops—AI can handle the complexities of modern software engineering. The significance of this development in AI history cannot be overstated; it is the first real-world application where AI agents are consistently performing high-level, multi-step professional labor at scale.

    As we move forward, the focus will likely shift from the capabilities of the AI to the governance of its output. The long-term impact will be a world where software is more abundant, more personalized, and faster to iterate than ever before. For developers, the message is clear: the future of coding is not just about writing syntax, but about mastering the art of the "agentic mission." In the coming months, watch for deeper integrations between these IDEs and cloud infrastructure providers as the industry moves toward a fully automated "Prompt-to-Production" pipeline.

    This content is intended for informational purposes only and represents analysis of current AI developments.

    TokenRing AI delivers enterprise-grade solutions for multi-agent AI workflow orchestration, AI-powered development tools, and seamless remote collaboration platforms.
    For more information, visit https://www.tokenring.ai/.

  • The Autodev Revolution: How Devin and GitHub Copilot Workspace Redefined the Engineering Lifecycle

    The Autodev Revolution: How Devin and GitHub Copilot Workspace Redefined the Engineering Lifecycle

    As of early 2026, the software engineering landscape has undergone its most radical transformation since the invention of the high-level programming language. The "Autodev" revolution—a shift from AI that merely suggests code to AI that autonomously builds, tests, and deploys software—has moved from experimental beta tests to the core of the global tech stack. At the center of this shift are two divergent philosophies: the integrated agentic assistant, epitomized by GitHub Copilot Workspace, and the parallel autonomous engineer, pioneered by Cognition AI’s Devin.

    This evolution has fundamentally altered the role of the human developer. No longer relegated to syntax and boilerplate, engineers have transitioned into "Architects of Agents," orchestrating fleets of AI entities that handle the heavy lifting of legacy migrations, security patching, and feature implementation. As we enter the second week of January 2026, the data is clear: organizations that have embraced these autonomous workflows are reporting productivity gains that were once thought to be the stuff of science fiction.

    The Architectural Divide: Agents vs. Assistants

    The technical maturation of these tools in 2025 has solidified two distinct approaches to AI-assisted development. GitHub, owned by Microsoft (NASDAQ: MSFT), has evolved Copilot Workspace into a "Copilot-native" environment. Leveraging the GPT-5-Codex architecture, the 2026 version of Copilot Workspace features a dedicated "Agent Mode." This allows the AI to not only suggest lines of code but to navigate entire repositories, execute terminal commands, and fix its own compilation errors iteratively. Its integration with the Model Context Protocol (MCP) allows it to pull live data from Jira and Slack, ensuring that the code it writes is contextually aware of business requirements and team discussions.

    In contrast, Devin 2.0, the flagship product from Cognition AI, operates as a "virtual teammate" rather than an extension of the editor. Following its 2025 acquisition of the agentic IDE startup Windsurf, Devin now features "Interactive Planning," a system where the AI generates a multi-step technical roadmap for a complex task before writing a single line of code. While Copilot Workspace excels at the "Human-in-the-Loop" (HITL) model—where a developer guides the AI through a task—Devin is designed for "Goal-Oriented Autonomy." A developer can assign Devin a high-level goal, such as "Migrate this microservice from Python 3.8 to 3.12 and update all dependencies," and the agent will work independently in a cloud-based sandbox until the task is complete.

    The technical gap between these models is narrowing, but their use cases remain distinct. Copilot Workspace has become the standard for daily feature development, where its "Copilot Vision" feature—released in late 2025—can transform a UI mockup directly into a working frontend scaffold. Devin, meanwhile, has dominated the "maintenance chore" market. On the SWE-bench Verified leaderboard, Devin 2.0 recently achieved a 67% PR merge rate, a significant leap from the mid-30s seen in 2024, proving its capability to handle long-tail engineering tasks without constant human supervision.

    Initial reactions from the AI research community have been overwhelmingly positive, though cautious. Experts note that while the "Autodev" tools have solved the "blank page" problem, they have introduced a new challenge: "Architectural Drift." Without a human developer deeply understanding every line of code, some fear that codebases could become brittle over time. However, the efficiency gains—such as Nubank’s reported 12x faster code migration in late 2025—have made the adoption of these tools an economic imperative for most enterprises.

    The Corporate Arms Race and Market Disruption

    The rise of autonomous development has triggered a massive strategic realignment among tech giants. Microsoft (NASDAQ: MSFT) remains the market leader by volume, recently surpassing 20 million Copilot users. By deeply embedding Workspace into the GitHub ecosystem, Microsoft has created a "sticky" environment that makes it difficult for competitors to displace them. However, Alphabet (NASDAQ: GOOGL) has responded with "Antigravity," a specialized IDE within the Google Cloud ecosystem designed specifically for orchestrating multi-agent systems to build complex microservices.

    The competitive pressure has also forced Amazon (NASDAQ: AMZN) to pivot its AWS CodeWhisperer into "Amazon Q Developer Agents," focusing heavily on the DevOps and deployment pipeline. This has created a fragmented market where startups like Cognition AI and Augment Code are forced to compete on specialized "Architectural Intelligence." To stay competitive, Cognition AI slashed its pricing in mid-2025, bringing the entry-level Devin subscription down to $20/month, effectively democratizing access to autonomous engineering for small startups and individual contractors.

    This shift has significantly disrupted the traditional "Junior Developer" hiring pipeline. Many entry-level tasks, such as writing unit tests, documentation, and basic CRUD (Create, Read, Update, Delete) operations, are now handled entirely by AI. Startups that once required a team of ten engineers to build an MVP are now launching with just two senior developers and a fleet of Devin agents. This has forced educational institutions and coding bootcamps to radically overhaul their curricula, shifting focus from syntax and logic to system design, AI orchestration, and security auditing.

    Strategic advantages are now being measured by "Contextual Depth." Companies that can provide the AI with the most comprehensive view of their internal documentation, legacy code, and business logic are seeing the highest ROI. This has led to a surge in demand for enterprise-grade AI infrastructure that can safely index private data without leaking it to the underlying model providers, a niche that Augment Code and Anthropic’s "Claude Code" terminal agent have aggressively pursued throughout 2025.

    The Broader Significance of the Autodev Era

    The "Autodev" revolution is more than just a productivity tool; it represents a fundamental shift in the AI landscape toward "Agentic Workflows." Unlike the "Chatbot Era" of 2023-2024, where AI was a passive recipient of prompts, the tools of 2026 are proactive. They monitor repositories for bugs, suggest performance optimizations before a human even notices a slowdown, and can even "self-heal" broken CI/CD pipelines. This mirrors the transition in the automotive industry from driver-assist features to full self-driving capabilities.

    However, this rapid advancement has raised significant concerns regarding technical debt and security. As AI agents generate code at an unprecedented rate, the volume of code that needs to be maintained has exploded. There is a growing risk of "AI-generated spaghetti code," where the logic is technically correct but so complex or idiosyncratic that it becomes impossible for a human to audit. Furthermore, the "prompt injection" attacks of 2024 have evolved into "agent hijacking," where malicious actors attempt to trick autonomous developers into injecting backdoors into production codebases.

    Comparing this to previous milestones, the Autodev revolution is being viewed as the "GPT-3 moment" for software engineering. Just as GPT-3 proved that LLMs could handle general language tasks, Devin and Copilot Workspace have proven that AI can handle the full lifecycle of a software project. This has profound implications for the global economy, as the cost of building and maintaining software—the "tax" on innovation—is beginning to plummet. We are seeing a "Cambrian Explosion" of niche software products that were previously too expensive to develop.

    The impact on the workforce remains the most debated topic. While senior developers have become more powerful than ever, the "Junior Developer Gap" remains a looming crisis. If the next generation of engineers does not learn the fundamentals because AI handles them, the industry may face a talent shortage in the 2030s when the current senior architects retire. Organizations are now experimenting with "AI-Human Pairing" roles, where junior devs are tasked with auditing AI-generated plans as a way to learn the ropes.

    Future Horizons: Self-Healing Systems and AGI-Lite

    Looking toward the end of 2026 and into 2027, the next frontier for Autodev is "Self-Healing Infrastructure." We are already seeing early prototypes of systems that can detect a production outage, trace the bug to a specific commit, write a fix, test it in a staging environment, and deploy it—all within seconds and without human intervention. This "Closed-Loop Engineering" would effectively eliminate downtime for many web services, moving us closer to the ideal of 100% system availability.

    Another emerging trend is the "Personalized Developer Agent." Experts predict that within the next 18 months, developers will train their own local models that learn their specific coding style, preferred libraries, and architectural quirks. This would allow for a level of synergy between human and AI that goes beyond what is possible with generic models like GPT-5. We are also seeing the rise of "Prompt-to-App" platforms like Bolt.new and Lovable, which allow non-technical founders to build complex applications by simply describing them, potentially bypassing the traditional IDE entirely for many use cases.

    The primary challenge that remains is "Verification at Scale." As the volume of code grows, we need AI agents that are as good at formal verification and security auditing as they are at writing code. Researchers are currently focusing on "Red-Teaming Agents"—AI systems whose sole job is to find flaws in the code written by other AI agents. The winner of the Autodev race will likely be the company that can provide the highest "Trust Score" for its autonomous output.

    Conclusion: The New Baseline for Software Production

    The Autodev revolution has fundamentally reset the expectations for what a single developer, or a small team, can achieve. By January 2026, the distinction between a "programmer" and an "architect" has largely vanished; to be a developer today is to be a manager of intelligent agents. GitHub Copilot Workspace has successfully democratized agentic workflows for the masses, while Devin has pushed the boundaries of what autonomous systems can handle in the enterprise.

    This development will likely be remembered as the moment software engineering moved from a craft of manual labor to a discipline of high-level orchestration. The long-term impact is a world where software is more abundant, more reliable, and more tailored to individual needs than ever before. However, the responsibility for safety and architectural integrity has never been higher for the humans at the helm.

    In the coming weeks, keep a close eye on the "Open Source Autodev" movement. Projects like OpenHands (formerly OpenDevin) are gaining significant traction, promising to bring Devin-level autonomy to the open-source community without the proprietary lock-in of the major tech giants. As the barriers to entry continue to fall, the next great software breakthrough could come from a single person working with a fleet of autonomous agents in a garage, just as it did in the early days of the PC revolution.

    This content is intended for informational purposes only and represents analysis of current AI developments.

    TokenRing AI delivers enterprise-grade solutions for multi-agent AI workflow orchestration, AI-powered development tools, and seamless remote collaboration platforms. For more information, visit https://www.tokenring.ai/.

  • The World’s First Autonomous AI Software Engineer: Devin Now Produces 25% of Cognition’s Code

    The World’s First Autonomous AI Software Engineer: Devin Now Produces 25% of Cognition’s Code

    In a landmark shift for the software development industry, Cognition AI has revealed that its autonomous AI software engineer, Devin, is now responsible for producing 25% of the company’s own internal pull requests. This milestone marks a transition for the technology from a viral prototype to a functional, high-capacity digital employee. By late 2025, the "Devins" operating within Cognition are no longer just experimental tools; they are integrated teammates capable of planning, executing, and deploying complex software projects with minimal human oversight.

    The announcement comes as the AI industry moves beyond simple code-completion assistants toward fully autonomous agents. Cognition’s CEO, Scott Wu, recently confirmed that the company's 15-person engineering team now effectively manages a "fleet" of Devins, with the ambitious goal of having the AI handle 50% of all internal code production by the end of the year. This development has sent shockwaves through Silicon Valley, signaling a fundamental change in how software is built, maintained, and scaled in the age of generative intelligence.

    Technical Mastery: From Sandbox to Production

    Devin’s core technical advantage lies in its ability to reason over long horizons and execute thousands of sequential decisions. Unlike traditional LLM-based assistants that provide snippets of code, Devin operates within a secure, sandboxed environment equipped with its own shell, code editor, and web browser. This allows the agent to search for documentation, learn unfamiliar APIs, and debug its own errors in real-time. A key breakthrough in 2025 was the introduction of "Interactive Planning," a feature that allows human engineers to collaborate on a high-level roadmap before Devin begins the execution phase, ensuring that the AI’s logic aligns with architectural goals.

    On the industry-standard SWE-bench—a rigorous test of an AI’s ability to solve real-world GitHub issues—Devin’s performance has seen exponential growth. While its initial release in early 2024 stunned the community with a 13.86% unassisted success rate, the late 2025 iteration leverages the SWE-1.5 "Fast Agent Model." Powered by specialized hardware from Cerebras Systems, this model can process up to 950 tokens per second, allowing Devin to "think" and iterate 13 times faster than previous frontier models. This speed, combined with the integration of advanced reasoning models like Claude 3.7 Sonnet, has pushed the agent's problem-solving capabilities into a territory where it can resolve complex, multi-file bugs that previously required hours of human intervention.

    Industry experts have noted that Devin’s "Confidence Scores" have been a game-changer for enterprise adoption. By flagging its own tasks as Green, Yellow, or Red based on the likelihood of success, the AI allows human supervisors to focus only on the most complex edge cases. This "agent-native" approach differs fundamentally from the autocomplete models of the past, as Devin maintains a persistent state and a "DeepWiki" intelligence of the entire codebase, allowing it to understand how a change in one module might ripple through an entire microservices architecture.

    The Battle for the AI-Native IDE

    The success of Devin has ignited a fierce competitive landscape among tech giants and specialized startups. Cognition’s valuation recently soared to $10.2 billion following a $400 million Series C round led by Founders Fund, positioning it as the primary challenger to established players. The company’s strategic acquisition of the agentic IDE Windsurf in July 2025 further solidified its market position, doubling its annual recurring revenue (ARR) to over $150 million as it integrates autonomous capabilities directly into the developer's workflow.

    Major tech incumbents are responding with their own "agentic" pivots. Microsoft (NASDAQ: MSFT), which pioneered the space with GitHub Copilot, has launched Copilot Workspace to offer similar end-to-end autonomy. Meanwhile, Alphabet (NASDAQ: GOOGL) has introduced "Antigravity," a dedicated IDE designed specifically for autonomous agents, and Amazon (NASDAQ: AMZN) has deployed Amazon Transform to handle large-scale legacy migrations for AWS customers. The entry of Meta Platforms (NASDAQ: META) into the space—following its multi-billion dollar acquisition of Manus AI—suggests that the race to own the "AI Engineer" category is now a top priority for every major hyperscaler.

    Enterprise adoption is also scaling rapidly beyond the tech sector. Financial giants like Goldman Sachs (NYSE: GS) and Citigroup (NYSE: C) have begun rolling out Devin to their internal development teams. These institutions are using the AI to automate tedious ETL (Extract, Transform, Load) migrations and security patching, allowing their human engineers to focus on high-level system design and financial modeling. This shift is turning software development from a labor-intensive "bricklaying" process into an architectural discipline, where the human’s role is to direct and audit the work of AI agents.

    A Paradigm Shift in the Global AI Landscape

    The broader significance of Devin’s 25% pull request milestone cannot be overstated. It represents the first concrete proof that an AI-first company can significantly reduce its reliance on human labor for core technical tasks. This trend is part of a larger movement toward "agentic workflows," where AI is no longer a chatbot but a participant in the workforce. Comparisons are already being drawn to the "AlphaGo moment" for software engineering; just as AI mastered complex games, it is now mastering the complex, creative, and often messy world of production-grade code.

    However, this rapid advancement brings significant concerns regarding the future of the junior developer role. If an AI can handle 25% to 50% of a company’s pull requests, the traditional "entry-level" tasks used to train new engineers—such as bug fixes and small feature additions—may disappear. This creates a potential "seniority gap," where the industry struggles to cultivate the next generation of human architects. Furthermore, the ethical implications of autonomous code deployment remain a hot topic, with critics pointing to the risks of AI-generated vulnerabilities being introduced into critical infrastructure at machine speed.

    Despite these concerns, the efficiency gains are undeniable. The ability for a small 15-person team at Cognition to perform like a 100-person engineering department suggests a future where startups can remain lean for much longer, and where the "billion-dollar one-person company" becomes a statistical possibility. This democratization of high-end engineering capability could lead to an explosion of new software products and services that were previously too expensive or complex to build.

    The Road to 50% and Beyond

    Looking ahead, Cognition is focused on reaching its 50% internal PR target by the end of 2025. This will require Devin to move beyond routine tasks and into the realm of complex architectural decisions and system-wide refactoring. Near-term developments are expected to include "Multi-Agent Orchestration," where different Devins specialized in frontend, backend, and DevOps work together in a synchronized "squad" to build entire platforms from scratch without any human code input.

    The long-term vision for Cognition and its competitors is the creation of a "Self-Healing Codebase." In this scenario, AI agents would continuously monitor production environments, identify performance bottlenecks or security flaws, and autonomously write and deploy patches before a human is even aware of the issue. Challenges remain, particularly in the areas of "hallucination management" in large-scale systems and the high compute costs associated with running thousands of autonomous agents simultaneously. However, as hardware specialized for agentic reasoning—like that from Cerebras—becomes more accessible, these barriers are expected to fall.

    Experts predict that by 2027, the role of a "Software Engineer" will have evolved into that of an "AI Orchestrator." The focus will shift from syntax and logic to system requirements, security auditing, and ethical oversight. As Devin and its peers continue to climb the ladder of autonomy, the very definition of "writing code" is being rewritten.

    A New Era of Engineering

    The emergence of Devin as a productive member of the Cognition team marks a definitive turning point in the history of artificial intelligence. It is the moment where AI moved from assisting humans to acting on their behalf. The fact that a quarter of a leading AI company’s codebase is now authored by an agent is a testament to the technology’s maturity and its potential to redefine the global economy’s digital foundations.

    As we move into 2026, the industry will be watching closely to see if other enterprises can replicate Cognition’s success. The key takeaways from this development are clear: autonomy is the new frontier, the "agent-native" IDE is the new battlefield, and the speed of software innovation is about to accelerate by orders of magnitude. For the tech industry, the message is simple: the AI colleague has arrived, and it is already hard at work.

    This content is intended for informational purposes only and represents analysis of current AI developments.

    TokenRing AI delivers enterprise-grade solutions for multi-agent AI workflow orchestration, AI-powered development tools, and seamless remote collaboration platforms.
    For more information, visit https://www.tokenring.ai/.