Good Morning! Here's what I have for you in today's newsletter:
OpenAI launches dots, always-on ChatGPT agents
OpenAI releases GPT-6.1 Sol, near-Astra coding at 1/5 of the price
Perplexity brings scheduled and event-triggered Automations to Computer
Learn five chat habits to prompt GPT-6.1 Sol the right way
4 new AI tools worth trying today
AI MODELS
OpenAI released GPT-6.1 Sol, an upgrade to GPT-6 Sol that gets close to GPT-6 Astra on coding, and computer use at 1/5 the token price.
Pricing stays at $2 per million input tokens and $10 per million output tokens. Cached input drops to $0.10 per million, a 95% discount on standard input.
On DeepSWE v1.1, it matches Astra at about 1/5 of the cost. On the OSWorld 2.0 computer-use test, it finishes 2.1 points behind Astra at roughly 1/7 of the cost.
It’s available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users through the API. It isn’t in regular Chat yet.
If you use Astra for coding agents or long automations, try Sol on a few of the same tasks and compare the results. The cheaper cached input could make a difference when agents repeatedly read the same codebase or documents. Astra still leads on the hardest problems, so test before moving everything over.
AI TOOLS
Perplexity launched Automations in Computer. You can now schedule recurring tasks or have the agent start when something happens in Slack, Gmail, Outlook, Linear, or GitHub.
You can set conditions for each trigger. A Gmail automation could run only when a specific sender asks for a decision, rather than on every new email.
Each run can use Computer’s memory, saved skills, connected apps, and results from previous runs, so you don’t have to give it the same instructions each time.
Automations replace Scheduled Tasks. Create them from Computer’s omnibar or the Automations panel. Credits are charged only when a task runs.
You could have Computer prepare a daily report, watch a Linear project for updates, or respond to emails from a particular client. Once you’ve set it up, you won’t need to start each task yourself. Watching for triggers is free; you pay credits when Computer carries out the work.
AI AGENTS
OpenAI introduced dots, always-on agents inside ChatGPT powered by GPT-6 Astra. Each dot has its own cloud computer and browser, with access to 4,000+ apps through ChatGPT plugins.
A dot keeps working between conversations, whether it’s booking a table or handling a longer project. You can message or call it in ChatGPT, Slack, and Teams without explaining everything again.
You choose what it can do on its own, when it needs your approval, and what it must never do. You can open its cloud computer anytime to see what it’s doing.
Access to your personal laptop is off by default. You can enable it in the ChatGPT desktop app, where you also create your first dot before using it on mobile.
Once you’ve set the rules, you can close the chat and leave your dot working. Access is rolling out to Pro and Business Premium users outside the EEA, Switzerland, and the UK. Enterprise admins can enable it for their teams. For the first month, conversations with a dot won’t count toward ChatGPT usage limits.
HOW TO AI
GPT-6.1 Sol now thinks before every reply. There's no "off" setting anymore.
That change can explain why prompts that worked on the old Sol model suddenly feel slower or use more of your allowance.
Step 1: Find the reasoning settings
Beneath the message box, click the model and reasoning control, then click Advanced: Light, Faster, Power, Smarter, Ultra. Power is the default.
The developer docs use different names for the same levels: low, medium, high, xhigh, max.
Step 2: Tell it what you need in five parts
GOAL: [what you want back, specifically]
CONTEXT: [only the facts it needs, not everything you have]
RULES: [what it must not do]
DONE MEANS: [how you'll know the answer is complete]
STOP AND ASK IF: [when it should check with you instead of guessing]
You don't need all five every time. A quick question may only need a goal and some context.
Step 3: Match the reasoning level to the task
Light: quick questions, simple rewrites. Power (default): most everyday work. Smarter: ambiguous or multi-step tasks. Ultra: when getting it right matters more than getting it fast.
Step 4: Tell it what to do when information is missing
Only use facts from what I've given you. If something isn't in there, say "not enough information" instead of guessing.On a deliberately hard factuality test, this same change dropped GPT-6.1 Sol's error rate from 11.4% to 7.7% compared to the old Sol model, at the lowest reasoning level.
Step 5: There's no "off" switch anymore
If you used "none" for fast answers with no reasoning, that's gone. The lowest setting now is Light, which still reasons briefly. A familiar prompt taking longer than before usually means this, not something broken.
P.S. You can access all the AI courses and trainings (including the full version of this one), prompts and workflows if you upgrade.
Meta added Monitor to Muse Code. It watches CI runs, builds, and logs in the background without making model calls, then wakes the coding agent when a check fails or something needs fixing.
Cognition opened Devin to ChatGPT subscribers. Plus and Pro users can sign in with their existing plan, with all OpenAI model usage in Devin Cloud, Desktop, and CLI counting against their ChatGPT quota.
Nous Research added Sign in with ChatGPT to Nous Portal. Built with OpenAI, the feature lets you run Hermes Agent on your ChatGPT plan and view and manage usage in ChatGPT settings.
📊 Cursor: Added /visualize to the Agents Window. It analyzes data and creates charts and diagrams directly in the conversation.
📝 Obsidian: Version 1.14.3 early access adds a Group menu for Kanban columns in Bases, macOS Quick Look previews, and a Quick Switcher that’s five times faster.
🔍 Lightreel: Describe the creator you’re looking for, and it searches millions of TikToks and Reels to find matching influencers and UGC creators.
🎙️ Inception: Released Mercury Voice, a diffusion LLM for voice agents with more than 2× lower latency. Enterprise customers can get access through sales.
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
✓ $500+ worth of AI tools