We felt this pain hard as an agency. So, we stopped guessing and built our own dashboard If you're running an AI agency or a hybrid team in 2026, you already know this problem. Your engineers are in Slack. Your AI agents are running quietly in the background. Somewhere between the two, visibility just disappears. We hit this at reStrucAI ourselves. Between interns, AI engineers, and a handful of AI agents actually doing client work, we couldn't answer a simple question: who — or what — is working on what, right now. Humans were tracked in one tool. Agents weren't tracked anywhere. We were mostly just trusting it was fine. "What gets measured, grows." Easy to say. We had nothing measuring anything. So, we stopped guessing and built our own operating system for a hybrid workforce. One dashboard. Human employees and AI employees, side by side, on the same screen. Who's online. What they're working on. How long a task is actually taking. Whether it's a person or an agent doing the work. If you're managing any team that's part human, part AI — this is the visibility gap almost nobody's built for yet. We built it to fix our own blind spot first. Now we're opening it up. Here's exactly what we packed into it 👇 https://youtu.be/BlYnoZre1aU?si=mNZDK_UN99ssiATo

I saw a comparison on YouTube and found it interesting - The author evaluates AI performance based on several strategic and economic frameworks, independent of specific brand or video content, while considering the roles of models like Fable 5 and ChatGPT 5.6 Sol: • Capability Tiers (Manager vs. Worker): This concept distinguishes between models suited for high-level reasoning, strategic planning, and creative direction (the "Manager," often associated with Fable 5) and those optimized for efficient, high-volume execution of code and routine tasks (the "Worker," often associated with ChatGPT 5.6 Sol). • Economic Efficiency: Practitioners analyze the balance between cost and performance by looking at: • Unit Economics: The total cost per completed request or project. • Token Efficiency: A model's ability to achieve results without unnecessary verbosity or inefficient "over-engineering." • Operational Reliability: Evaluating how a model's safety guardrails impact its utility. Strict refusal filters can enhance safety but may impede programmatic automation, whereas more permissive models often prove more reliable for high-velocity API workflows. • Speed vs. Latency Metrics: Distinguishing between median latency (predictable, typical speed) and mean latency (the mathematical average). High variance in latency can disrupt agentic loops, even if a model is fast on average. • Task-Appropriateness: The strategy of matching model sophistication to task complexity. Using an expensive, top-tier model for simple, stateless requests is considered inefficient compared to leveraging models specifically tuned for cost-effective execution What has been your experience ? I am yet to use both to compare. Although I feel ChatGPT uses token more efficiently.

One of the most impressive real-world AI implementations I’ve seen recently comes from the Government of Alberta. Using Claude Code with autonomous AI agents, they: - 🔍 Scanned 466 million lines of code across government systems in about 20 hours. - 🛡️ Identified security vulnerabilities, generated fixes, wrote missing tests, and even modernized legacy applications—with human review before deployment. - 🤖 Built specialized AI agents for continuous security reviews, red-team testing, blue-team validation, code quality, and documentation. The biggest takeaway isn’t just the technology—it’s the workflow. They didn’t replace engineers. They built AI agents that handle repetitive analysis, documentation, testing, and remediation, allowing experts to focus on validation and decision-making. This is a great example of AI moving beyond chatbots to becoming an operational teammate for large, complex organizations. 💭 Question for the community: If you could deploy AI agents inside one department of your organization today, where would you start—IT, finance, customer support, compliance, or somewhere else?

The rebuilt ChatGPT platform emphasises a unified, multi-functional workspace designed for both everyday users and advanced productivity needs. Key capabilities include: 1. Advanced Workflow Integration 1.1. Work Tab: Introduces a systematic planning-and-action loop that automates complex tasks. It allows users to define specific inputs, integrate various plugins, and manage projects in a single, cohesive environment. 1.2. Unified Application: The new desktop app consolidates several previously standalone tools, merging Web Browsing, Codex functionality, and the new Work interface into one streamlined hub. 2. Intelligent Document Synthesis & Creation 2.1. Document Processing: The models are highly adept at synthesising vast amounts of data from disparate sources—such as PDFs, Excel sheets, and reports—to generate comprehensive documents, spreadsheets, or presentations. 2.2. Sites Feature: A powerful paid-plan capability that allows users to generate, host, and refine professional-looking web applications or websites directly within the chat window, providing a clean alternative to traditional file sharing. 3. Next-Generation Model Family (GPT-5.6) 3.1. Performance: The new models—Saul, Terra, and Luna—offer top-tier performance benchmarks. Specifically, Saul delivers capabilities comparable to leading models like Fable but at a significantly reduced operational cost. 3.2. Ultra Mode: For high-complexity tasks, the Ultra setting triggers a multi-agent system that coordinates at least four independent AI agents to work in tandem, allowing for deeper, more reliable task execution. 4. Enhanced Interaction & Utility 4.1. Real-Time Voice Mode: Features significantly improved latency, supporting fluid, live translation between languages. The mode is designed to handle natural conversational flow, including the ability for the AI to speak over or alongside the user during interactions. 4.2. Plugin Ecosystem: With over 1,500 plugins, the platform supports highly diverse use cases, now bolstered by the ability to import specific 'skills' —a format for custom instructions that allows for personalised writing voices and specialised task management.

The timing on this is almost too perfect. Let me connect the dots 👇 The news: Z.ai (formerly Zhipu AI) released GLM-5.2 with full MIT-licensed open weights. The benchmarks: → #2 on Code Arena, behind only Fable 5 — ahead of Claude Opus 4.7 AND 4.8 in thinking mode.→ Within ~1 point of Opus 4.8 on agentic coding.→ Beats GPT-5.5 on long-horizon coding benchmarks — at roughly 1/6th the cost.→ Ranks #1 among ALL open-source models. It's a 753B mixture-of-experts model that only activates ~40B parameters per query, so it punches at frontier level without frontier compute. MIT license means you can download it, fine-tune it, and run it commercially with basically zero restrictions. Now here's why the timing matters: Three days ago, the US government pulled Fable 5 and Mythos 5 offline for all foreign nationals. So for any developer or business outside the US, the most capable Claude models just vanished overnight. And right into that gap drops GLM-5.2 — open weights, self-hostable, frontier-class coding, a fraction of the cost. That's not a coincidence in vibes, it's a coincidence in consequence: when access to closed models gets politically fragile, open weights stop being the budget option and become the strategic one. My take for builders and operators: This is the exact argument I made in my Fable shutdown post, now proven in real time. Model access is a geopolitical variable. The thing that makes open weights powerful isn't just price — it's that nobody can switch them off. You download GLM-5.2 once, it's yours. No directive, no export control, no API ban takes it away. For anyone building client work: a model this capable, this cheap, that you can self-host, is a genuine option for production now — not a toy. The "open models can't keep up" excuse is dead. #2 on coding, behind only a model the government just banned, says it plainly. The honest caveat — and it's important: If you use Z.ai's hosted API, you're subject to China's National Intelligence Law, which can compel data sharing. For anything sensitive — healthcare, client data, regulated work — that's a real risk. BUT: the MIT open weights let you self-host and sidestep that entirely. So the move for serious use isn't "call their API," it's "run the weights yourself." Know the difference. It matters.