Frontier models, AI agents, and voice automation, tested in the open and written up by the people building with them.
63
Dispatches
07
Desks
12
Live Topics
Watts spoke to IBM engineers in 1969 and argued with Arthur C. Clarke in print about electronic intelligence, HAL and the Turing test. He also got one thing wrong. Here is what he actually said, and how to tell it from what people write in his name.
GPT-6.1 Terra cut prices 50 percent. NorthCUA claims open-source parity at 397B. Both are true, and both mislead, because the number that decides your budget sits three lines further down the same chart.
Read article
Seven memory products shipped in sixteen days. The most useful number published all month came from Alibaba's own engineers, and it measures the gap every vendor benchmark hides.
Read article
One week in October captured the whole problem. A telecom started issuing employee numbers to its AI agents while a developer marketplace shipped a backdoored build of the tool those agents are built with.
Read article
Canada is adding $500 million to regional AI support. The opportunity is bigger than another model announcement: useful deployment, paying customers and productivity gains that outlast a funded pilot.
Read article
Henry Shevlin's move to Google DeepMind brings questions about AI minds and human relationships into the lab. The practical challenge is separating what a model can do, what it claims to feel and how people should be protected.
Read article
Hinton's contributions to AI are beyond serious dispute. His extinction forecasts deserve a different kind of scrutiny: explicit assumptions, testable mechanisms, and evidence that distinguishes capability from control.
Read article
A fabricated homicide tip submitted during an AI evaluation shows why agent safety needs more than accurate text: it needs firm boundaries between reading, drafting and acting.
Read article
Gemini 4 Argon leads several enterprise evaluations, but coding results are mixed. We review the published scores, independent evidence, pricing and gated rollout.
Read article
Meta's latest Muse Spark pairs long-horizon reasoning with sharply different API prices. The useful story is in its evaluation methods, not a single leaderboard crown.
Read article
Both personal agents can keep working in the cloud. Their real differences are where your context lives, what they can change, and how much supervision the job deserves.
Read article
A precise reading of the new Super Intelligence order, and an Intueo design perspective on America.gov's ambitious approach to public services.
Read article
Canada's AI opportunity spans enterprise models, creative tools, and physical systems. Here is how to evaluate that ecosystem without confusing a brand with a deployment guarantee.
Read article
Anthropic's latest release shifts the argument from maximum intelligence to affordable results. The crucial number is not the token price, but the cost of work that actually passes review.
Read article
Europe's AI content icons make disclosure more visible. The harder question is whether a label tells audiences what changed, who checked it, and who remains responsible.
Read article
Google's new conversational avatars could make digital interactions easier to follow. They also sharpen the distinction between a system that looks attentive and one that deserves confidence.
Read article
As frontier labs use AI to build future models, oversight needs its own scorecard. Coverage, review timing and independent verification matter more than another reassuring percentage.
Read article
AI infrastructure is meeting the slower clock of electricity grids. The challenge is not simply generating more energy, but delivering power in the right place, at the right time, without shifting the bill to everyone else.
Read article
OpenAI says a swarm of 10,000 AI agents produced a proof about the Navier–Stokes equations in 88 hours, followed by a Lean formalization. The result could mark a turning point for AI-assisted mathematics—but a company announcement and a machine-checked artifact are not the same as broad mathematical acceptance. Here is what was claimed, what the problem asks, and what must happen next.
Read article
Anthropic’s Claude Fable 5.1 and OpenAI’s GPT-6 Astra arrive within days of each other, but the meaningful contest is no longer who writes the best answer. Both models are built for long-running agentic work across code, computers, research, and professional software. Their differences—in cost, computer use, scientific reasoning, safeguards, and deployment—show what enterprises should evaluate before trusting a model to act.
Read article
The United Nations is using AI to make textbooks accessible to children with disabilities, detect methane leaks from space, and accelerate tuberculosis screening where specialists are scarce. These projects offer a useful counterweight to the AI hype cycle: measurable outcomes, narrowly defined tasks, community participation, and human accountability. They also show why deploying AI for human rights requires more than good intentions.
Read article
Nvidia is reportedly paying Poolside $6 billion for a non-exclusive license to the startup’s Model Factory and offering jobs to 109 of its employees, while investing another $1 billion in the company. The unusual deal is not simply a talent acquisition or a bet on coding models. It is an attempt to industrialize how American open-weight AI gets built—and to answer the speed, efficiency, and accessibility of China’s model ecosystem.
Read article
Two releases in August 2026 reveal where AI is heading next. A vendor-neutral plugin format gave agents a portable way to acquire skills, while managed agent payments gave them a controlled way to buy digital services. Together they point to a new software layer: agents that can discover capabilities, install them, and pay for what they use. Here is what changed, what remains risky, and why the next AI platform may look less like an app store and more like a governed machine economy.
Read article
AI agents crossed the line from demo to coworker this year, and the whole industry is now racing to answer a question nobody asked during the pilot phase: who is this agent, what is it allowed to do, and can we prove what it did? Adoption is real but shallow, interoperability standards finally landed at the Linux Foundation, and regulators and security researchers are converging on identity and governance as the make-or-break layer. Here is what the primary sources actually say, and what it means for anyone deploying agents.
Read article
Sixty years after the first chatbot fooled its own creator's secretary, the psychology is clear: humans instinctively treat machines that talk like social beings. That reflex makes AI feel supportive and easy to use — and it is also what makes emotional over-reliance, misplaced trust, and manipulative design possible. Here is what the peer-reviewed evidence actually says, and how to build AI that respects the reflex instead of exploiting it.
Read article
The fear that AI 'makes us dumber' is too simple. The research points somewhere more useful: AI reliably extends what a thinking person can do, and it just as reliably erodes the skills of a person who stops thinking. The difference is not the tool — it is the workflow. Here is what the evidence on cognitive offloading, automation bias, and critical thinking actually shows, and a practical protocol for keeping human judgment in the loop.
Read article
On July 6, Anthropic published one of the strangest interpretability results yet: its Fable models spontaneously developed a silent internal workspace — a place where concepts are held, weighed, and sometimes deliberately withheld from the output. The tool that found it, the Jacobian lens, can read those hidden concepts out loud. We break down what J-space actually is, why it matters for anyone deploying agents, and we built a playable audit console so you can try reading a model's mind yourself.
Read article
OpenAI's gpt-realtime models process voice as voice — no transcription relay, no robotic pauses, just one model listening and speaking in a single stream. We have been integrating it into our voice AI agents, and the difference on a live phone call is not subtle. Here is how the technology works, what changed for our callers, and why the model is only half of the story.
Read article
The defining AI trend of mid-2026 is not a new model — it is a new job description. Agents are moving out of chat sidebars and into actual roles: Microsoft is shipping sales and service agents grounded in CRM data, Slack is wiring agents across business systems, and Gartner says up to $234 billion in SaaS spend could shift toward outcome-based agentic platforms. Here is our field report on what is real, what is struggling, and what the 43% who cannot measure value are getting wrong.
Read article
We are winding down development of the Sage platform as a standalone product. Instead, we are focusing on what our clients actually asked us for: custom AI environments built around each business — bespoke apps and a private company platform where your team talks to and manages its own agents. Here is what is changing and why.
Read article
A new GPT just shipped, and the frontier is still moving. But the bigger story of 2026 is happening underneath it: open-weight Chinese models like GLM 5.2 are landing within a hair of Opus-class quality at a fifth of the price. We have been running GLM 5.2 in production at Intueo, and it is genuinely good. Here is an honest, long-form look at where the frontier labs stand, where the Chinese models stand, what the cyberattack headlines actually mean for you, and the one question every buyer keeps getting stuck on: where does my data live?
Read article
If you want an AI employee instead of a clever bot that forgets you every morning, the foundation matters more than the model. Claude Code and Codex are remarkable and amnesiac: every session starts from zero. Letta Code does not. It is a stateful agent operating system that persists across sessions, learns your business, writes its own skills, and carries its memory across machines and even across model generations. This is a sourced, head-to-head look at Letta Code versus Claude Code, why people are buying dedicated mini PCs just to keep agents running, why Letta Code is the best agent OS we have found, and how we turn it into stateful AI employees your business can actually depend on.
Read article
For two years, every chatbot greeted you like a stranger. In 2026 that quietly ended. ChatGPT, Claude, and Gemini have all shipped persistent memory, each with a wildly different philosophy about what your AI should remember and who controls it. We have been living in this future for a while, so here is an honest, sourced map of the memory wars, where each giant landed, and the line that separates a chatbot that remembers a few facts from a stateful agent you can actually employ.
Read article
AI voice agents are answering and making millions of business calls a day, and the law has quietly caught up. AI-generated voices are now regulated like robocalls, consent rules already apply, and disclosure requirements are tightening. This is a clear, sourced guide to what is actually legal in 2026, the mistakes that turn a helpful voice agent into a liability, and why we believe compliance has to be designed into a voice agent from the first line, not bolted on after launch.
Read article
Most AI voice receptionists forget every caller the moment they hang up. Here's the real problem we hit at Intueo, how we built one that actually remembers using our stateful Sage agents, and why it's the part that matters most. (Currently in private testing, available on request for a demo.)
Read article
An MIT study found that 95% of enterprise generative-AI pilots delivered no measurable bottom-line impact, even as companies poured tens of billions into them. The headline reads like an indictment of AI. It is not. It is an indictment of how AI gets deployed. Here is a sourced look at why pilots die in the gap between demo and production, what the successful minority actually do, and why a stateful, integrated, operated approach is the difference between a science project and an employee.
Read article
We spent the last few days testing Mythos, and it was unlike anything else in the field. Then the most impressive model we have ever used became the center of a national-security storm: cyber-capability flags, curtailed access, an investor reportedly uneasy in advance, governance debate stretching to India, and a stinging rebuke from AI godfather Geoffrey Hinton. Here is what actually happened, grounded in the record, and why it matters for everyone who builds on this technology.
Read article
Two announcements in the same week pushed autonomous AI agents from helpful assistants to actual buyers. Visa integrated with ChatGPT so agents can purchase on their own at Visa merchants, while Coinbase is leaning into AI-driven portfolio trading. This is the moment agentic AI stops talking and starts spending.
Read article
For all the excitement around what AI can build, the same technology is being turned to harmful ends. A surge in AI-generated "nudify" tools is fueling bullying, especially among kids, while broader fears about misinformation and AI in elections keep growing. This is the conversation the industry cannot afford to skip.
Read article
Anthropic just launched Claude Fable 5 and Claude Mythos 5, the first generally available Mythos-class models. We break down every major claim, the benchmarks, the science, the safeguards, and the pricing, and we share what we are seeing first-hand running Fable 5 inside the Intueo platform.
Read article
OpenAI has been steadily turning ChatGPT into something that can plan, schedule, and carry out multi-step work on its own. We think that is great news, even for a platform like ours that has been building in this direction for a while.
Read article
At WWDC 2026, Apple unveiled a complete overhaul of Siri, rebuilt on Apple Intelligence and a new partnership with Google. Here is what changed, and why it matters for anyone who cares about conversational AI.
Read article
After months of quietly running pilot projects alongside a handful of generous early partners, we are excited to announce that Intueo's three enterprise service pillars are now open to everyone. Here is what we built, who helped us prove it works, and what you can do with it today.
Read article
From agents that finally do real work to the rise of persistent memory and browser-native AR, the AI landscape is moving fast. Here is a brief, practical roundup of the trends we are watching most closely this season.
Read article
We built a multi-agent orchestration system where AI agents could team up, assign leaders, and communicate through A2A protocols. After months of testing, we scrapped it entirely. Here is why shared memory turned out to be the better architecture.
Read article
Most AI tools today are generic assistants. Intueo is something different. It is an operating system for building, running, and managing a team of specialized AI agents, each with its own personality, expertise, persistent memory, and toolset. Here is how it works.
Read article
The enterprise call center is undergoing its biggest transformation in decades. As voice AI platforms now handle millions of calls per month, rigid keypad menus are being replaced by natural conversation. Here is what that shift means for your business.
Read article
We were thrilled to represent Intueo Labs at the second annual Web Summit Vancouver, joining over 20,000 attendees from more than 100 countries to showcase our vision for agentic AI that transforms how businesses operate.
Read article
As Mental Health Awareness Month approaches, we examine how generative AI and language models are providing accessible support to millions who cannot access traditional therapy, backed by emerging research showing promising results.
Read article
The release of Opus 4.6 arrived without the typical theatrical demos. But for those building autonomous infrastructure, this is the most significant model update in over a year. A look at why operational dependability matters more than benchmark headlines.
Read article
How an AI receptionist can keep your schedule clean, reduce no-shows, and connect calls directly to your calendar and CRM. Discover the difference between conversational bots and agentic voice systems.
Read article
Claude Opus 4.5 blends frontier AI performance with cost efficiency. Discover what's new, benchmark results, and how it unlocks intelligent automation across business workflows.
Read article
Google has quietly entered a new phase of its AI roadmap with Gemini 3. This post breaks down the new reasoning capabilities, tool integration, and what it actually means if you build with AI.
Read article
How to tell if your bots and voice AI are actually working. This post breaks down the core metrics that matter for AI in customer service and voice automation.
Read article
How things are actually shifting right now: From chatbots to agentic AI, breaking classic IVR, and the rise of stateful agents as the new default.
Read article
Are your teams drowning in busywork? Here are 5 clear signals that it's time to implement AI automation to scale your operations and reclaim your time.
Read article
Why most AI agents fail at complex tasks, and how 'Stateful AI' is solving the problem by remembering context, preferences, and history just like a human employee.
Read article
A complete step-by-step guide to implementing an AI receptionist that handles calls, books appointments, and provides 24/7 customer service without the overhead.
Read article
Traditional IVR menus frustrate callers. Learn how modern conversational AI creates intelligent phone experiences that feel like talking to a concierge, not navigating a maze.
Read article
Voice is the most natural way to communicate. Discover why leading businesses are shifting from text-based chatbots to human-quality Voice AI for superior customer experiences.
Read article
Chatbots answer questions. AI employees complete work. Learn why stateful, tool-connected AI employees deliver exponentially more value than traditional conversational interfaces.
Read articleWeekly insights on AI automation and voice technology, straight to your inbox.
