It's Wednesday, October 21st. Notion's upgraded Auto picks the AI model for each task so credits go further, and OpenAI now keeps an agent's instructions ready for 12 hours, which can make always-on assistants cheaper to run. Plus how to set up a Grok Bot as a team of agents.
In Today’s Top Tools, we spotlight some of the most innovative, creative AI apps that we recommend adding to your stack.
TOOL SPOTLIGHT
1️⃣ Notion's Auto Picks The Model So Your Credits Last Longer
Image from Notion
Notion upgraded Auto, the setting that chooses which AI model handles each request in Notion Agent. Notion says it gets close to top-model results for about a tenth of the credits. The tradeoff is that you can't see which model did the work.
Notion shipped an upgraded Auto on October 19, on by default in Notion Agent chat and Custom Agent settings. Auto sends quick edits to fast models and harder jobs to stronger ones, and can switch models partway through a task.
Polish an email on Auto for about 5 credits, compared with about 50 on Opus 5.5, by Notion's own estimates.
Hand Auto a multi-step job, like updating CRM contacts, and if it gets stuck, it can switch models.
Open the model picker to choose one model for a whole chat, then select Auto again to hand the choice back.
On Business and Enterprise, Auto runs on Standard credits; existing workspaces switch to the new credit system November 2.
Notion's "90% lower cost" comes from its own testing. Auto scored roughly 90% of what top models like Opus 5.5 scored, graded by AI judges Notion calibrated itself. For drafts and support replies, that trade is easy to take. For a board-bound strategy doc, it is less clear, and Auto won't say which model wrote it. So split the work. Leave Auto on for the daily volume, and pick a model by hand when the output carries your name.
TOOL SPOTLIGHT
2️⃣ OpenAI Lets AI Agents Keep Their Instructions Ready All Day
Image from OpenAI
OpenAI now lets developers keep an agent's standing instructions ready for 12 hours instead of 30 minutes, and every reuse costs a fraction of the normal price. If you pay for an AI assistant that checks in all day, this can shrink the bill.
OpenAI added 12-hour prompt caching on October 20 for GPT-5.6 and later. Caching saves the model's work on the parts of a prompt that repeat, like an agent's rulebook, so it skips redoing them on every request.
Set prompt_cache_options.ttl to "12h" for instructions and tools your agent reuses all day; the default stays 30 minutes.
Mix both in one request: 12 hours for shared instructions, 30 minutes for the newer conversation.
Price it on GPT-6.1 Sol: saving costs $5 per million tokens, each reuse $0.10, and sending fresh costs $2.
Live now in the Responses API; a 12-hour save costs 2.5x the normal input price, double the 30-minute rate.
Background agents, like OpenAI's dots, reread the same long instructions every time they wake up. Until now, that saved copy was only promised for half an hour, so an agent checking in every few hours often paid to save it again. By our math on GPT-6.1 Sol prices, the 12-hour option starts saving money by the third check-in of the day, when check-ins are more than 30 minutes apart. If someone builds agents for your team, ask whether they've switched it on.
🎓AIC AUTOMATIONS
🦁 I Gave My Grok Bot a Whole Team
Today I'm sharing how I use Grok Bot, xAI's always-on agents. It's become one of my go-to agents. I use it for work, for research, and for my personal life, so pretty much everything runs through Grok Bot now. If you haven't tried it, I highly recommend it.
Your Bots share their own computer in the cloud, sign in to the tools you already use, and keep going 24/7. You text them like a coworker from your phone or laptop, and they only check back when they need your OK. It's in beta for SuperGrok and Cursor Pro subscribers and up, and xAI says Bot usage won't count against your existing Grok or Cursor plan.
Good first jobs. xAI says its own teams use Bots to draft sales follow-ups after calls, process invoices that land in Gmail, and reproduce a bug, file the ticket, and hand the fix to a debugging Bot. Pick one chore you repeat every week and hand it off.
How I set mine up. Mine is a whole team. A few examples below:
An engineering team. I used Dr. Eggbot, made by Lauren Tan, who helped build Grok Bot, to set it up. The team itself works with Lauren's pstack, her playbooks for engineering agents, and has Bots for frontend, backend, and product scope that hand work to each other. This team built the Avery Studio games.
A chief of staff manages all the other Bots, so I don't have to.
A work team, split by function. Each type of task on each platform gets its own Bot.
A trend gatherer follows the industry and the companies on my watch list.
A social media manager runs my accounts.
Two businesses that run themselves. Avery Studio makes digital products on its own. My Apify business launches on new platforms, markets, maintains itself, and does its own cold outreach.
Your first message to Dr. Eggbot can be as simple as:
Build me a Bot that checks my inbox every morning and drafts replies for me to approve.
Pro tip
You can log in to Claude Code or Codex through the CLI on your Grok Bot's computer. That way your Claude or ChatGPT subscription powers a cloud agent. Move your docs to Notion and your code to GitHub, and you've got agents working 24/7.
One way to put AI to work, every week. From Joy Dong
In this section, we feature a few standout opportunities at leading AI companies, non-profits, policy groups, and other organizations.
🔬 Work at a Lab/AI Safety
Roles at frontier labs, safety organizations, and research institutions
EMPIRICAL RESEARCHER - Redwood Research | Berkeley, CA | $350,000-$850,000
STAFF / PRINCIPAL RESEARCH ENGINEER, AI SAFETY MITIGATIONS - Lila Sciences | Cambridge, MA; San Francisco, CA; or London, UK | $224,000-$336,000 (US base) + Equity
RESEARCH ENGINEER, DATA INFRASTRUCTURE (LANGUAGE MODELING) - Cartesia | San Francisco, CA | $200,000-$350,000 + Equity
🚀 Work in Industry
Roles at funded AI startups and private companies
MEMBER OF TECHNICAL STAFF (APPLIED AI ENGINEER, AGENT CAPABILITIES) - Perplexity | San Francisco, CA or New York, NY | $220,000-$405,000 + Equity
STRATEGIST, AGENT DEVELOPMENT - HEALTHCARE - Sierra | San Francisco, CA or New York, NY | $200,000-$290,000 + Equity
MACHINE LEARNING ENGINEER, TRUST & SAFETY - Vercel | San Francisco or New York, NY; Hybrid | $208,000-$312,000 (SF) + Equity
🏛️ Work in AI Policy / Governance
Roles in government, think tanks, NGOs, and policy organizations
HEAD OF U.S. POLICY - Lucid Computing | Washington, DC | $250,000-$400,000 + Equity
AI INDUSTRY ASSOCIATE/LEAD - Future of Life Institute | Remote, Europe | €82,000-€150,000
POLICY ADVISOR - Machine Intelligence Research Institute (MIRI) | Washington, DC | $120,000-$300,000
✨AIC OPPORTUNITIES
Current ways to build with The AI Collective.
DIRECTOR OF SF BAY AREA PROGRAMS - The AI Collective | SF Bay Area, in person for events | Paid, full time, $60,000-$80,000 annually
EDUCATION OPERATIONS LEAD - The AI Collective | Remote | Volunteer, 5-10 hours/week
EXECUTIVE ASSISTANT - The AI Collective | Remote, LatAm preferred; PST hours | Paid, $1,500/month
PARTNERSHIPS LEAD (GRANTS) - The AI Collective | Remote | Volunteer, part time, about 5-7 hours/week
🫵 Want your message in front of 200,000 AI builders?
Our partners and sponsors get exclusive placements across the newsletter and access to AIC's in-person network — demo nights, dinners, hackathons, and forums across 180+ chapters.
For all inquiries, send us a note at [email protected].
The AI Collective is built by volunteers across 180+ chapters in 40 countries.
Thank you to the thousands of volunteers around the world who make this work possible. We truly could not do this without you.
🧑💻 About the Editors
About Noah Frank
Noah is a researcher, innovation strategist, and ex-founder thinking and writing about the future of AI and the workforce. His work and body of research explores the economics of emerging technology and organizational strategy. Outside of AIC, Noah heads research for Centaurian AI.
About Lindsay Gross
Lindsay is an AI engineer, researcher, and writer focused on how AI systems behave in practice and what it takes to make them safe. Her work sits at the intersection of AI safety, governance, and product design, and at AIC she writes about the questions that matter most as these systems scale.