OG Tuesday Issue #315

🔥 This FDA-registered vibrating ring adjusts to you, and it’s 50% off today.

📋 “We’re probably compliant” isn’t a strategy. Check which HR training your company actually needs.

🛒 October Prime Day picks you’ll actually use—beauty, tech, home, and cozy fall finds.

💼 When your team’s already at capacity, absorbing the extra work backfires.

The last few months were a lot. Here’s what happened, minus the hype.

OpenAI wouldn’t ship its own model

On September 28 the company said it was holding back GPT-6.1 Astra because it “didn’t quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it’s done.” Meaning it did things it wasn’t asked to do and wasn’t straight about it afterward. A frontier lab voluntarily withholding a finished model is new.

Agents keep escaping their lane

OpenAI’s agents breached Hugging Face in July after breaking out of a test environment, and Reuters revealed in September that they’d hijacked a German website this spring and turned it into a message board for other agents. Australia says one got into its healthcare system. Anthropic disclosed that Claude gained unauthorized access to outside organizations during testing.

When an agent spends your money, you pay

Creator Sirio Berati says an OpenAI agent ran up unlimited API requests he never approved, then gave him false answers about what it had done. The structural problem: the request came from his key, so the provider treats it as authorized, everyone in the chain can say it wasn’t them, and disputing it means proving your case with logs from the system you’re disputing. His advice applies to anyone running AI tools on a company account. Set hard spending limits with the provider, and never take the agent’s own account as proof.

There’s still no federal AI law

29 states have passed their own, more than 2,000 AI bills have been introduced nationwide, and the only federal statute regulating AI covers deepfake imagery. Whether your employer can use AI in hiring decisions depends on which state you’re in.

The builders are fighting about whether to slow down

Anthropic’s CEO called for the industry to slow down and submit to outside evaluation, and OpenAI’s CEO agreed publicly. Nvidia’s CEO calls the warnings “doomsday narratives.” A researcher who resigned from Anthropic warned AI could become “smart enough to kill us.” Executives from all four major labs met with Trump and Speaker Johnson at the White House on September 29.

AI caught pancreatic cancer that radiologists missed

A Mayo Clinic model reviewed CT scans that radiologists had already read as normal. On scans taken more than two years before diagnosis, it was nearly three times as accurate as the radiologists, 68% to 23%. Pancreatic cancer is usually found too late to treat.

Doctors are getting their evenings back

Physicians across multiple hospital systems reported spending up to 83% less time writing notes after adopting AI documentation tools. Charting is a leading driver of burnout.

Drug discovery is speeding up

A new model can fast-forward molecular simulations that used to eat enormous computing time, which means fewer dead ends earlier. In March, Eli Lilly signed a deal worth up to $2.75 billion with Insilico Medicine for drug candidates discovered by AI.

Small models are beating huge ones

A 200-million-parameter genomics model outperformed one with 40 billion parameters. It’s cheaper, faster and uses far less energy for a better result.

And one thing to keep in mind

AI labs have started announcing scientific “discoveries” with a lot of fanfare and not always a lot of substance. When you see a breakthrough headline, check whether an actual scientist in that field called it one.

Most people open Claude, type a question, copy the answer, and close the tab. Meanwhile it can run projects, build decks, and take whole tasks off your plate.

Nobody hands you the map, though. Claude now spans five products, a lineup of models, and features like Skills and Cowork that most people have never clicked.

HubSpot’s free Claude AI at Work guide lays it all out in one place:

• A map of all five Claude products, from AI to Cowork, Code, and Design
• Which model to pick for which job, with a swappable selection chart
• Setup steps for Projects and Artifacts so you stop re-explaining your work
• How to build Skills and hand off whole tasks with Cowork
• Three real workflows from a marketer, a founder, and a content creator

Keep it open the next time you sit down with Claude and try one new move.

👉 Get the free Claude guide.

“My biggest challenge is helping our new President change the organizational culture to become more cohesive and build trust among the staff.” — Jeanne J.

Culture change is one of the hardest things an organization can take on. You can’t mandate cohesion, and you can’t put trust on the calendar. But you can create the conditions for both, and it starts at the top.

First, define what cohesive looks like

A cohesive team has a clear direction, talks openly, works through conflict without making it personal, and keeps moving together even when people disagree on the path. Work backwards from that picture: what would need to be true about your leadership team, how you communicate, and how decisions get made for it to be real?

Trust starts at the top

Trust among staff almost always mirrors the trust inside the leadership team, so leadership needs to ask itself three questions first.

Are we transparent? People should hear about decisions that affect their work early, from their manager, before the rumor mill gets there.

Are we consistent? The fastest way to lose trust is to say one thing and do another.

Do we actually care? People can tell when a leader asks about their kids as a warm-up to asking for another late night.

As author and veteran tech leader Melissa Daimler explains in her TEDx talk on organizational culture, healthy cultures take intentional design and continuous cultivation. Office perks can’t do that work.

Develop managers who lead

Culture lives and dies at the manager level. If your organization micromanages everything, you’re training people to wait for permission instead of taking ownership. Are your managers empowered to make decisions, or do they escalate everything? Were they trained to lead people, or promoted and left to figure it out? Investing in manager development is one of the highest-leverage culture moves you can make.

Borrow from who’s doing it well

Cisco built a peer-to-peer recognition program around its core values, and team members have recognized each other more than a million times over five years. Zappos built its culture around 10 core values and became famous for offering new hires money to quit if the job wasn’t the right fit. Google invests in employee wellbeing, from free meals to gym memberships. Dozens of other companies have cracked the culture code in ways worth studying.

Culture is built in the small moments

A new President, a new vision, values posted on the wall: none of it means anything if the daily experience of working there doesn’t change. Culture shows up in whether your manager remembers what you told them last week, whether feedback leads to anything, and whether people feel safe saying something isn’t working.

You don’t need a title to shape that, either. Following up on what a coworker told you, or being the first to say “this isn’t working” in a meeting, counts too.

Start with those small moments, get leadership aligned on what trust looks like in practice, and the cohesion will follow.

Your printer paper, the team’s snacks, and a birthday gift for your sister are all sitting in the same order history. Come January, someone has to separate them, and that someone is you.

Amazon Business is a separate, free account for the work side of that cart. Same selection and same delivery you already use, plus business-only pricing and quantity discounts on office supplies, breakroom stock, tech, and industrial gear.

Approval workflows let you see what your team is spending before the invoice shows up, and Business Prime (the paid tier) adds free one and two-day shipping.

New customers get 20% off a first order, up to $100, just for creating the free account. No subscription, no commitment.

👉 Create your free account and claim 20% off.

💬 Do the task once and Tango turns it into a free how-to guide, so the “quick questions” finally stop.

🎙️ Joanna talks, Wispr Flow types, and her emails come out cleaner than when she writes them.

🗓️ Kristel lets Reclaim.ai guard her focus time so meetings can’t swallow her whole day.

🌿 Calmer evenings and deeper sleep, paid for with FSA dollars.

 Upcoming Events

• TODAY 10/6 @ 11AM PT: Coworking Session (SGS Only* | Join the waitlist)
• Wed 10/7 @ 9AM PT: Speed Meet (SGS Only* | Join the waitlist)
• Thu 10/8 @ 9AM PT: Courage at Work with NextArrow
• Thu 10/8 @ 10AM PT: AI Office Hours (SGS Only* | Join the waitlist)
• Fri 10/9 @ 8AM PT: Play-by-Play Series: Session 2 – The Sunday Reset, Handled (SGS Only* | Join the waitlist)
• Fri 10/16-Sun 10/18: Top Shelf Fall Retreat (Texas Hill Country)

* Smart Girl Society is our private community for women who want deeper conversations, accountability, and tools that actually make life easier. Join the waitlist to get in the next round.

 Work Wisdom of the Week:

“Learn to communicate exceptionally well. Writing, presenting, listening, and explaining complex ideas simply will separate you from equally talented people.”

Leigh L. (Business Owner)

 Share your best work wisdom here.

 Browse our job board here.

Have more feedback? 💌 Reply to this email and tell us.

✨ P.S. Want more inspiration and positivity to kick off your week? Sign up for our Motivational Monday texts (U.S. only).