Good Morning! Here's what I have for you in today's newsletter:
Google introduces Gemini 3.8, its best reasoning and coding model yet
Anthropic open-sources its blueprint for shopping and merchant bots
Meta releases Muse Spark 1.3 with better agentic performance and usability
Learn the exact fixes Anthropic recommends for prompting Fable 5.1
4 new AI tools worth trying today
AI MODELS
Google introduced Gemini 3.8, its best reasoning and coding model yet, releasing two new variants, Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, building on the momentum of 3.7 Flash from just three weeks earlier by leveraging long-running agentic loops.
Long-running agentic loops let the model sustain multi-step reasoning across extended tasks, rather than resetting its approach with each new step.
Two distinct variants ship together, with Flash Cyber positioned as a more specialized option alongside the standard Flash release.
The three-week gap since 3.7 Flash's release shows a notably fast iteration pace for a model family update of this scale.
Shipping a new model just three weeks after the last one is fast, even for AI. Releasing two variants at once targets different workloads directly, instead of forcing everything through one general model.
AI AGENTS
Anthropic open-sourced Claude Commerce Agents, a blueprint for building shopping and merchant agents, with reference implementations spanning retail, travel, telecom, and entertainment.
The blueprint covers both sides of a transaction, shopping agents that act on a buyer's behalf and merchant agents that operate on the seller's side.
Reference implementations across four distinct industries give developers a working starting point rather than a purely theoretical framework.
Making this open source means any developer can inspect, modify, or build directly on Anthropic's own reference architecture instead of building commerce agent logic from scratch.
Commerce is one of the clearest real jobs for agents, since it requires actually searching, comparing, and buying, not just talking. Open-sourcing this across four industries gives developers a real starting point, instead of everyone solving the same problem alone.
AI TOOLS
Meta released Muse Spark 1.3, improving performance on agentic and coding tasks with a focus on real-world usability, sustaining longer-horizon work, collaborating more actively with users, and using about 20% fewer tool calls and 25% fewer tokens than Muse Spark 1.2 in internal comparisons.
The model now asks clarifying questions, flags when it's stuck, and confirms before consequential actions, rather than proceeding silently through uncertainty.
Better calibration on its own limits means it's less likely to hallucinate an outcome it hasn't actually verified.
The reduction in tool calls and tokens compared to the previous version suggests genuine efficiency gains, not just raw capability improvements.
An agent that flags uncertainty and checks before acting is solving a different problem than one that's just smarter. It's earning trust, and using fewer tool calls and tokens for the same results also saves real money, not just a nicer demo.
HOW TO AI
Your old Fable 5 prompts mostly still work on Fable 5.1. Mostly. A handful of behavioral differences show up often enough that Anthropic wrote an entire page about them.
Step 1: Change your effort levels
Effort level names don't mean the same thing across models. Start at high, the default, then test the others against something you actually care about.
Run this same request at medium effort instead of high, and tell me if anything meaningful changedStep 2: Ask for progress updates
Fable 5.1 writes fewer updates during long tasks than earlier models did. Add this if you want visibility:
Before you start, say in a line what you're about to do. Brief updates while you work help me follow along. Close with a short recap that stands on its own.Step 3: Kill "mannered writing"
Fable 5.1's output can run dense, swapping a direct statement for metaphor and flourish. Anthropic's own team named this anti-pattern.
Please remove all mannered writing.
Step 4: Tell it exactly when to format
Fable 5.1 already leans toward plain paragraphs. An old anti-formatting rule now fights the model's default instead of correcting it.
Use lists and bullet points when asked to, or when the content is multifaceted enough that they help with clarity. If I explicitly ask for minimal formatting, always respond in plain paragraphs.Step 5: Make it finish the whole task
On open-ended tasks, Fable 5.1 sometimes stops early, describing what it would do next instead of doing it.
I'm not watching this in real time and can't answer questions mid-task, so asking "want me to?" will just block things. For anything reversible that follows from what I already asked, go ahead without checking first.
P.S. You can access all the AI trainings (including the full version of this one), prompts and workflows if you upgrade.
Anthropic added the ability for Claude to use your computer in the background in Claude Cowork and Claude Code, clicking, typing, and opening apps on your desktop while you work on something else.
Google introduced flexible usage limits in Gemini Notebook, rolling out deferred artifact generation and shifting limit resets to every 5 hours instead of daily.
Perplexity open-sourced Lily, the local inference engine built for hybrid compute in Perplexity Computer, specialized for Qwen3.6-35B-A3B on Apple silicon so on-device compute doesn't bottleneck tasks.
🎬 Imagine: now supports up to 14 references per video, images, voices, and character references, tagged directly with @.
☁️ Cursor: cloud agents can now run on your own infrastructure, including auto-scaling machine pools with access to internal services.
🔧 Runway Dev MCP: connect to the Runway developer platform from inside your coding agent to build, debug, and manage without leaving it.
💻 OpenCode: the open source coding agent, free, model-agnostic, works with OpenAI, Claude, Gemini, and more, 195,000+ GitHub stars.
THAT’S IT FOR TODAY
Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!
See you tomorrow :)
- Dr. Alvaro Cintas
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial