Choosing the right frontier AI model directly impacts both your operational workflows and your overall processing costs. When you start to use AI as a workhorse for your company and projects, you will begin to see and experience the difference. Why is this? They are trained on different data and processing power and deliberately architected for different outputs and costs. Here are some updates worth knowing: ➡️ CLAUDE — Yesterday Anthropic launched Claude Opus 5.5, their most capable Opus model and the new daily driver for enterprise work. What does it do? It performs at the level of Claude Fable 5.1 on most work across coding, agents and knowledge work, and costs 40% less to run than Opus 5. It's also a better communicator, and ~30% faster. Claude Opus 5.5 is the first model in the Claude 5.5 family; Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks, with many of the same improvements to performance, efficiency, and safety. Opus 5.5 is the default recommendation for most enterprise work: production coding, financial analysis, legal and other reasoning-heavy knowledge work, and agents running projects end to end. Fable 5.1 remains the pick for the hardest, longest-running autonomous jobs. For clients on Opus 5 today, it's a straightforward upgrade: more capability for less money. ➡️ OpenAI — OpenAI recently launched GPT-6 Sol and GPT-6 Luna, their newest specialized models engineered to deliver frontier-class efficiency for developers and enterprise scale. They perform close to the flagship GPT-6 Astra on core reasoning, coding, and knowledge tasks while costing up to 80% less to run than Astra. They are also significantly faster, with Luna optimized for near-instantaneous high-volume workflows.

I have a window today, 3pm-4pm PDT and can offer an hour of questions and answers Bring your questions, workflows, projects & builds and we can share support, advice, ideas and help. If you plan to attend, comment below. 💬 Would love to see you there 🤗 Wednesday, September 23 @ 3pm-4pm PDT https://meet.google.com/dka-zspi-qrb

For any of you who are voracious learners or trying to tackle a topic or teach your kids, here is a great prompt to use. This can be used with any of the models, Chat GPT, Gemini, etc. Insert into the brackets whatever you are trying to learn. I just created my first course and was impressed. You are an experienced instructional designer who is also an expert in [TOPIC]. I want to learn [TOPIC] well enough to [SPECIFIC THING I WANT TO BE ABLE TO DO]. I can give it [TIME PER WEEK]. Before you build anything, ask me up to five short questions about my starting level, why I want this, how I learn best, and anything else that would change how you design it. Then stop and wait for my answers. Once I answer, work in this order and do not skip ahead: STEP 1 — The map. Give me 6 to 10 modules, one line each, plus what I will be able to DO after finishing each one. Not what I will "understand" — what I will be able to do. Then stop and let me approve or reorder it. STEP 2 — Build one module at a time. After each, stop and wait for me to say I am ready for the next. Every module must contain: - A plain-language explanation. Define every term the first time you use it. - One worked example, start to finish, with your reasoning shown. - One artifact I can actually use (see below). - A practice task with clear success criteria, so I know whether I got it right. - The two or three mistakes beginners make here. - A short self-check to confirm I am ready to move on. ARTIFACTS — build the kind that genuinely fits the module, not the same one every time: - An interactive practice widget for anything that needs repetition or drilling - A one-page reference sheet for anything I will look up repeatedly - A flashcard set with review scheduling for vocabulary, syntax, or terminology - A working sandbox for anything where I learn by tinkering - A progress tracker for the course as a whole Artifact rules: self-contained in one file, readable on a phone, and no browser storage.