Aloha From Hawaii. The July 2026 Fishing Report.

The AI Hot Seat: Real Problems, Solved Live, On Stage.

This month we debuted something brand new: the AI Hot Seat. A two-hour session where we bring audience members up on stage and solve their real business problems live, in front of the whole room. No slides about what AI could do. Just AI doing it, in real time, with the people who would use it.

The stage was BEMA, the 107-year-old association that brings the world's leading baking companies together with the largest manufacturers of baking equipment. This year's convention landed on the Big Island of Hawaii, and roughly 300 leaders packed the room.

Here is what we built, live on stage. A brand-new website for a company that manufactures blades for the baking industry, built from scratch in front of the audience. A live dashboard for a company that sells manufacturing equipment, turning raw numbers into a tool their team could use immediately. And the one AI initiative that 25 brands in the room could implement today to grow sales and profit tomorrow.

None of it was pre-built. No staged demos. Everything created in the room, with the audience watching every step. Because AI is not a writing assistant. It is a real-time business tool.

The room reported back: 98.6% found the session valuable. 98.5% would attend again. 100% found it engaging, relevant, and actionable. The overall session rating was 9.83 out of 10. One attendee wrote: 'Up to date, fresh and disruptive. Loved the live examples and the potential.' Another said: 'Jon brought AI down to a real-world, actionable perspective with examples specific to our industry. Very relatable. I'm a slow adopter, and now I'm intrigued and interested in making it a bigger part of my life.' A third: 'Very valuable, up-to-date information that is actually actionable. It opened up my mind to how much more I can use AI in my life and business. AI is a tool, not to replace jobs, but to EXPAND them.'

Fable 5 Is Back. And Anthropic Just Opened a Window Into Claude's Mind.

Last month we covered a first: after Anthropic released Fable 5, its most powerful public model, the US government ordered access removed for any foreign national, and Anthropic pulled the model down entirely. This month, the story turned again. Fable 5 is back for Claude subscribers as a new government review framework takes shape.

But the bigger story this month is not the return of Fable 5. It is what Anthropic published on July 6: research describing a small, privileged internal space inside Claude that behaves like a functional workspace. They call it J-space, found with a tool called the Jacobian lens. It is a place for thoughts the model can hold in mind, use for multi-step reasoning, and sometimes reveal before they show up in the final answer.

Here is what is real, without the hype: this is not proof that Claude is conscious. Anthropic is careful to separate conscious access, meaning information a system can report and reason over, from subjective experience, which remains an open question. What it does give us is a more concrete way to see what a model is doing internally: where it privately notices it is being tested, or holds a goal it has not stated yet. For safety and trust, that visibility matters.

Jon's take: 'The interesting part is not whether it is alive. It is that we can finally look inside and check whether a model means what it says. That is a tool, not a headline.'

GPT-5.6 Is Almost Here. Meet Sol, Terra, and Luna.

OpenAI previewed the GPT-5.6 family on June 26, and broad availability is expected imminently. Like Fable 5, it launched first to a small group of government-approved partners under a safety review before reaching everyone. The pattern is clear: frontier models now ship through a government checkpoint.

The GPT-5.6 family introduces three named tiers. Sol is the flagship, with maximum reasoning and a new ultra mode that runs subagents on hard, multi-step work. Terra is the everyday workhorse, delivering near-flagship-class quality at about half the cost. Luna is fast and affordable, built for high-volume, low-latency tasks.

The naming finally makes sense: the number is the generation, and Sol, Terra, and Luna are capability tiers you pick by the job. For teams, that is the real shift. Stop asking which model is best. Start routing work to the right tier. A quick email response does not need Sol. A competitive analysis does. Matching the tool to the task is where the real efficiency gains live.

Our Current Favorites: The Five Tools We Keep Reaching For.

The AI landscape moves weekly, so here is where we are spending our time right now. These are the five tools we keep reaching for.

Claude CoWork and Design is our home base. CoWork for real, agentic work and Design for building branded pages, decks, and prototypes, including this very newsletter. If you are doing serious writing, analysis, or design work, Claude is where we start.

Manus for websites and dashboards. Astonishing for website and dashboard design, and a big part of what powered the live builds on stage at BEMA. If you need to build something functional from a description, Manus is the tool.

ChatGPT for image generation. The image generation is genuinely stunning now: clean text-in-image, fast edits, real production-quality output. If you create marketing materials, social content, or presentations, this is worth exploring today.

Gemini for video creation. Our go-to for AI video right now, and the quality and control keep pulling ahead of the competition.

ElevenLabs for voice and voice changing. Best-in-class voice generation and voice changing, the finishing touch on videos, demos, and localized content. If you are building multilingual content or need a professional voice layer on your AI-generated video, this is the tool.