Sponsored by
Hey friends 👋
Big week for anyone tracking the AI arms race. OpenAI didn't just ship a new model, it started merging its entire product lineup into one app that puts an agent in front of nearly a billion people. And a new survey shows the split between executives who think AI adoption is going great and the managers actually stuck doing the work by hand. Let's dig in.
GPT-5.6 launched, and honestly the model was the least interesting part
OpenAI released GPT-5.6 on Thursday, and yes, the benchmarks are good. The new lineup includes Sol as the flagship, Terra as the balanced option, and Luna as the cheap and fast one. They're reportedly beating Anthropic's Mythos and Fable on tests like Agent's Last Exam, and they're doing it for less money. Sol runs $5 per million input tokens and $30 for output, compared to $10 and $50 for Mythos and Fable.
But the model wasn't the headline. OpenAI used the same announcement to fold Codex into the main ChatGPT app, launch a new "ChatGPT Work" mode for agentic tasks, and roll out "ChatGPT Sites," which lets anyone describe a website in plain English and get a live, shareable link back. That last one quietly replaces tools like Lovable, Replit, and Vercel for a lot of casual use cases.
Here's the thing worth sitting with. OpenAI has about 5 million weekly Codex users and close to a billion ChatGPT users. This launch is the moment agents stop being a power-user feature and start showing up for basically everyone with a ChatGPT login, whether they asked for that or not. It's also worth noting the government reportedly asked OpenAI to delay broader release until agencies and companies could prepare for the model's ability to surface software vulnerabilities. That's not a small footnote. A model good enough to find security holes at scale is good enough to be misused at scale too.
AI moves faster than any single newsletter can cover. To actually stay current you'd need to read every major AI newsletter, listen to the AI podcasts that move thinking, watch what every AI lab ships, and scan public-company earnings calls, investor letters, and GitHub. Almost no one has time for that.
MavSource fixes this in 5 minutes a day. We aggregate all of those sources, summarize what matters, analyze the trends shaping AI — across both public and private markets, apps and infrastructure — and send it as one daily email with founder commentary on what the day's signal actually means.
No checking dozens of sources. No hype. One brief, every morning. Built for the people who can't afford to fall behind. Free.
CEOs think AI adoption is going great. Ask a manager and you'll hear something else
New survey data from document software company Nitro found a pretty stark gap. 85% of C-suite executives say they've deployed AI across their organization. Only 54% of managers say it's actually a priority for their team, and barely half say AI has touched their day-to-day workflows at all.
Of the managers whose teams have adopted AI, 37% are just copying and pasting into ChatGPT rather than using anything integrated. Only 12% say AI is genuinely built into how their documents move through the workflow. Meanwhile, 62% of managers say their employees are still spending six or more hours a week on manual document work, and nearly a third put that number above 11 hours.
"If AI isn't built into the workflows, work surfaces, and systems people use every day, with the outcome in mind, it doesn't matter how many tools you buy or how many tokens you burn," Nitro CEO Cormac Whelan told The Deep View.
This is the gap that never seems to close. Executives need to say "we're using AI" for the board and the shareholders. Managers need the tool to actually work inside the mess of existing systems, and most AI rollouts don't bother with that part. More than half of managers admit sensitive documents are ending up in public AI tools, and less than half say their company even has a clear, enforced policy. That's not an adoption problem, that's a governance problem wearing an adoption costume.
A few more things worth knowing
Anthropic launched a "reflection dashboard" in Claude's settings that shows you your usage patterns, common topics, and lets you set quiet hours and break nudges. It only works if memory is turned on, and it skips incognito chats and health data.
On the same day it rolled out that self-awareness feature, Anthropic also reset the 5-hour weekly rate limits for Fable users, clearly aimed at matching demand created by GPT-5.6.
Screen time trackers on phones taught us that just measuring a habit rarely changes it. Worth keeping in mind before anyone calls a usage dashboard a fix for AI overreliance.
That's what stood out to me today. Reply and tell me what caught your eye, I read everything.
Talk tomorrow,
Hatman 🎩