Hermes Agent NEW Super-App and DeepSeek v4 Catches Up To Opus 4.8?
Riley Brown · 2026-06-06
💡 Quick Take
1. Codeex is evolving into a full-fledged coding platform with its new "Sites" feature.
2. The "Sites" feature in Codeex allows for the creation and deployment of internal tools accessible only to your team.
3. Codeex can integrate with your Gmail and Slack to build personalized and relevant internal sites.
4. Codeex now has a dedicated "Sites" tab for managing all your generated internal tools.
5. Codeex is integrating database platforms like Convex, enabling agents to update databases from any chat.
6. Codeex now has a built-in plugin to render Swift (iOS) apps directly within the Codeex interface.
7. Cursor has launched "Canvases," a feature mirroring Codeex's "Sites" for creating and sharing internal tools.
8. Cursor is introducing "Profiles," aiming to create a GitHub-like platform for AI coding.
9. Both Codeex and Cursor are focused on making it easy for users to create internal websites and tools connected to company data.
10. Cursor's long-term vision appears to be a "GitHub for AI coding," enabling easy sharing and forking of AI-generated apps.
11. The initial internal-only release of features like Codeex Sites and Cursor Canvases is due to potential security risks with broader sharing.
12. Open-source AI models, particularly DeepSeek V4, are showing significant improvements and cost-effectiveness for agentic tasks.
13. Companies are switching from expensive proprietary models (Anthropic, OpenAI) to cheaper, high-performing open-source alternatives like DeepSeek V4.
14. The cost difference between DeepSeek V4 Pro and models like Anthropic's Opus 4.8 or OpenAI's GPT 5.5 is substantial, making open-source models far more viable for consumer agent products.
15. High costs of premium AI models make them inaccessible for many consumer-focused AI agent companies, driving a need for cheaper alternatives.
16. Anthropic's calls for a pause on AI development and regulation may be influenced by their IPO aspirations and competition from open-source models.
17. There are multiple incentives at play regarding AI development announcements, including IPO valuations, risk mitigation, and geopolitical competition (US vs. China).
18. Poke.com has become the first official AI agent approved to operate directly within Apple's iMessage.
19. Apple's iMessage approval for Poke.com signals a potential shift towards iMessage becoming a primary "super app" for general consumers.
20. Apple's upcoming WWDC is speculated to include announcements about an agent approval process for iMessage, potentially transforming SaaS into agents.
21. Hermes Desktop has launched, offering an open-source alternative to proprietary "super apps" with improved features and model flexibility.
22. Hermes allows users to switch between various AI models from different providers (Anthropic, OpenAI, DeepSeek, etc.).
23. Microsoft is releasing new AI updates, including the OpenClaw companion app integrated into Windows and a new "Dev Box" potentially capable of running local AI models.
24. Microsoft has announced its first set of "world-class frontier models" for AI.
📊 Detailed Explanation
1. Codeex is evolving into a full-fledged coding platform with its new "Sites" feature. This is a massive leap for Codeex, moving beyond just code generation to a complete environment where code can be built, deployed, and shared as functional tools.
2. The "Sites" feature in Codeex allows for the creation and deployment of internal tools accessible only to your team. This means you can prompt Codeex to build something like a dashboard, and it will automatically deploy it as a private web application that your team can access, enhancing collaboration and productivity.
3. Codeex can integrate with your Gmail and Slack to build personalized and relevant internal sites. By connecting to your existing communication and productivity tools, Codeex can pull in specific data (like important emails or Slack updates) to create highly tailored and useful internal applications for your team.
4. Codeex now has a dedicated "Sites" tab for managing all your generated internal tools. This organizational improvement makes it easy to find, access, and manage all the internal applications you've created within Codeex, streamlining workflow.
5. Codeex is integrating database platforms like Convex, enabling agents to update databases from any chat. This is a game-changer for workflow automation. You can now add tasks to a Kanban board (stored in Convex) by simply typing a command in *any* chat, even with a different AI agent, and it will update your Codeex-hosted tool. This creates a unified system for task management.
6. Codeex now has a built-in plugin to render Swift (iOS) apps directly within the Codeex interface. This is fantastic for iOS developers! Instead of just getting the code, you can see your app come to life in a side-by-side view within Codeex, making the development and testing process much more immediate and visual.
7. Cursor has launched "Canvases," a feature mirroring Codeex's "Sites" for creating and sharing internal tools. This shows a clear trend in the AI agent space: building platforms that allow users to easily create and deploy functional applications for internal use, fostering a new era of "vibe coding."
8. Cursor is introducing "Profiles," aiming to create a GitHub-like platform for AI coding. By allowing users to claim handles and connect their data, Cursor is laying the groundwork for a more public and collaborative ecosystem for AI-generated applications, similar to how GitHub functions for traditional code.
9. Both Codeex and Cursor are focused on making it easy for users to create internal websites and tools connected to company data. This is a core use case for AI agents, enabling companies to leverage their existing data to build highly relevant and personalized internal tools that can significantly boost efficiency.
10. Cursor's long-term vision appears to be a "GitHub for AI coding," enabling easy sharing and forking of AI-generated apps. The ultimate goal seems to be democratizing software creation and sharing in the AI era, allowing anyone to build, share, and contribute to AI-powered applications, much like the open-source movement on GitHub.
11. The initial internal-only release of features like Codeex Sites and Cursor Canvases is due to potential security risks with broader sharing. The ability to connect AI agents to company data is powerful but also risky. Limiting these features to internal use initially is a responsible step to prevent potential data breaches or misuse before more robust security measures are in place.
12. Open-source AI models, particularly DeepSeek V4, are showing significant improvements and cost-effectiveness for agentic tasks. This is a huge development! Open-source models are no longer just catching up; they're starting to outperform proprietary models in specific areas, especially when cost is a factor.
13. Companies are switching from expensive proprietary models (Anthropic, OpenAI) to cheaper, high-performing open-source alternatives like DeepSeek V4. The example of Lindy switching 100% of traffic to DeepSeek V4, saving millions and seeing performance increases, is a powerful testament to the viability of open-source AI for real-world applications.
14. The cost difference between DeepSeek V4 Pro and models like Anthropic's Opus 4.8 or OpenAI's GPT 5.5 is substantial, making open-source models far more viable for consumer agent products. We're talking about 23x to 27x cost savings! This is critical for the sustainability of consumer-facing AI agent products that need to be affordable for everyday users.
15. High costs of premium AI models make them inaccessible for many consumer-focused AI agent companies, driving a need for cheaper alternatives. The current pricing of top-tier models is simply too high for startups aiming for mass consumer adoption. The goal is to bring costs down to pennies per month, which open-source models are paving the way for.
16. Anthropic's calls for a pause on AI development and regulation may be influenced by their IPO aspirations and competition from open-source models. It's worth considering the timing and potential motivations behind such calls. If open-source models are becoming competitive and threatening their market share, advocating for regulation could be a strategic move to slow down that competition, especially as they prepare for an IPO.
17. There are multiple incentives at play regarding AI development announcements, including IPO valuations, risk mitigation, and geopolitical competition (US vs. China). When you see major announcements or statements from AI companies, it's important to look beyond the surface. Factors like investor expectations, genuine safety concerns, and national interests all play a role in shaping the narrative.
18. Poke.com has become the first official AI agent approved to operate directly within Apple's iMessage. This is a massive validation for AI agents and a significant step for Apple. It means AI can now be a first-class citizen within one of the most widely used messaging platforms.
19. Apple's iMessage approval for Poke.com signals a potential shift towards iMessage becoming a primary "super app" for general consumers. While power users might opt for dedicated AI platforms, for the average person, the familiar interface of iMessage could become the primary way they interact with AI, making it the ultimate consumer super app.
20. Apple's upcoming WWDC is speculated to include announcements about an agent approval process for iMessage, potentially transforming SaaS into agents. If this happens, it's revolutionary. It means any existing software-as-a-service (SaaS) could be re-imagined as an AI agent, accessible through iMessage, fundamentally changing how we interact with software.
21. Hermes Desktop has launched, offering an open-source alternative to proprietary "super apps" with improved features and model flexibility. For those who prefer open-source solutions and want more control, Hermes Desktop is a compelling option. It's seen as a superior alternative to OpenClaw, offering a more polished experience.
22. Hermes allows users to switch between various AI models from different providers (Anthropic, OpenAI, DeepSeek, etc.). This model agnosticism is a key advantage, giving users the freedom to choose the best model for their specific needs and budget, rather than being locked into one provider.
23. Microsoft is releasing new AI updates, including the OpenClaw companion app integrated into Windows and a new "Dev Box" potentially capable of running local AI models. Microsoft is finally making some serious moves in the AI space. The integration into Windows and the potential for local model execution on their Dev Box could be a big deal for Windows users.
24. Microsoft has announced its first set of "world-class frontier models" for AI. This is a significant announcement, as Microsoft has been somewhat behind in releasing its own cutting-edge models. Their performance will be crucial to watch, especially for Windows users and those in the enterprise space.
🎯 Expert Opinion
Wow, what a whirlwind week in the AI agent landscape! It's clear we're witnessing a rapid evolution, and the "super app" trend is really taking hold, but with some fascinating nuances. Codeex and Cursor are neck-and-neck, essentially creating internal development environments that are incredibly intuitive for non-coders. The "Sites" and "Canvases" features are brilliant because they democratize app creation for internal use, bridging the gap between an idea and a functional tool. This is the essence of "vibe coding" – making development accessible and collaborative. The integration with databases like Convex is particularly powerful, as it means these generated sites aren't just static displays; they're dynamic, interactive tools that can manage real data.
The push towards a "GitHub for AI coding" by Cursor is a bold and, I believe, inevitable move. We've revolutionized how software is *generated*, and now the focus is shifting to how it's *shared* and *evolved*. This will fundamentally change software development and distribution. However, the cautious approach of keeping these features internal-first is wise. The potential for misuse or security breaches when connecting AI agents to sensitive company data is immense. This phased rollout is a smart strategy to build trust and refine security protocols before opening the floodgates.
The cost-effectiveness of open-source models like DeepSeek V4 is a seismic shift. For years, the prohibitive cost of proprietary models has been a major bottleneck for consumer AI products. This price parity (or even superiority) for open-source alternatives is a massive win for innovation and accessibility. It means smaller companies and individual developers can now realistically build sophisticated AI agents without breaking the bank. This will undoubtedly lead to an explosion of new consumer-facing AI applications. The narrative around Anthropic's "pause" call is complex; while safety is paramount, the timing alongside their IPO plans is certainly noteworthy. It's a reminder that in the AI race, there are multiple competing agendas at play – genuine safety concerns, market positioning, and financial ambitions.
Apple's move with iMessage is a masterstroke. By officially approving Poke.com, they're signaling that messaging platforms are the future of user interaction for the masses. Forget complex dashboards; the most powerful AI agent might just be a text message away. This has massive implications for how businesses will engage with customers and how individuals will manage their digital lives. If WWDC confirms a broader agent approval process, it will be a paradigm shift, turning every SaaS into a potential iMessage agent. This is the kind of disruption that redefines entire industries.
Hermes Desktop is a crucial player in the open-source ecosystem. It provides a much-needed alternative for those who want flexibility and control over their AI models. Its ability to swap out models on the fly is a significant advantage, catering to a diverse range of needs and budgets. And finally, Microsoft's renewed focus on AI with their integrated companion app and potential for local model execution on Dev Boxes is a significant development for Windows users. While they've been playing catch-up, these announcements suggest they're serious about competing in the AI frontier, especially with their own "frontier models." The next few months are going to be incredibly exciting as these trends converge and mature.
Kanal: Riley Brown