Using AI to get off work at 5 pm instead of 6 pm is fun. Using AI to shave 4 hours per day off your work and grow your company or your role into something completely different... that is really really fun.

That is this week's deep dive. The gap between companies pulling ahead with AI and companies just "using AI" is growing, and it has nothing to do with your sector. I also want to hear from you, so there is a short reader survey at the bottom. Five questions. Takes two minutes.

This week: AI literacy and the widening gap, Opus 4.7 (already feels like old news), OpenAI's Codex doing its best Claude impression, and Claude Design taking a 7% bite out of Figma's stock price on day one.

Here is something I have noticed working with a lot of different businesses over the last 6 months: getting your business to be efficient with AI has almost nothing to do with what sector you are in.

It has everything to do with how open your mind is and how well you actually know what the tool can and can't do.

And the gap between companies using AI to its full potential and companies just "using AI" is actually growing. Fast.

6 months ago, "full potential" meant some decent prompting and seeing where AI could save you time here and there. Thirty minutes on an email. An hour on a report. Nice. Useful. But basically marginal.

That is not what full potential means anymore.

In Claude alone right now you have connectors, skills, chat, Cowork, Claude Code, scheduled tasks, artifacts, projects, Claude Design (more on that below), and more. These are not features that save you a few minutes on a task. These are tools that, together, can run entire workflows for you. The difference is not marginal. It is the difference between growing your company because of AI and your team getting off at 5 pm instead of 6 pm every day.

But here is the catch, and it is a real one. The activation energy is not small.

A lot of people approach AI hoping it is a do-all tool where they don't have to put in much effort. And honestly, a little effort does yield pretty good results. That is the trap. You save 30 minutes a day, you feel good about it, and you stop learning.

If you treat AI like learning Excel, the math changes completely. You go from saving 30 minutes a day to saving four hours a day. You go from shaving time off existing work to actually turning your business around.

It takes real time and real energy to learn these tools. I am not going to pretend otherwise. But the ROI is greater than any tool currently or previously available to businesses. (No shade to the internet or the spreadsheet.)

This is what AI literacy means. It is not knowing which model to pick. It is understanding the capabilities well enough to recognize when a workflow can be automated end to end versus when you should just prompt your way through it.

We spend a big chunk of our time at Bridge AI on exactly this. Not selling AI. Teaching people what the tools actually do so they can make real decisions about their own work. If that sounds useful, reply to this email.

Claude Opus 4.7 is here (and already feels like old news)

Anthropic shipped Opus 4.7 last Thursday. Big jumps on coding, much better vision (3x the image resolution), better instruction following, and a new "xhigh" effort level for really hard problems. Same pricing as 4.6 on APIs.

My take: Here is what gets me. This happened not even a week ago and it already feels like old news. That is how fast AI is moving right now. A model that would have been the biggest tech story of the year two years ago is now just... Thursday. If you are running a business and not paying attention to this pace, the gap I wrote about above is going to show up in your P&L. Read the migration guide before you just swap models though. Prompts written for 4.6 can sometimes behave differently because 4.7 follows instructions more literally.

READ THE ANTHROPIC ANNOUNCEMENT →

OpenAI's Codex update looks suspiciously like Claude

Also last Thursday (busy day), OpenAI dropped a major Codex update. Background computer use, in-app browser, image generation, memory, long-running automations, 90+ plugins, SSH to remote devboxes. The comparison writes itself.

My take: It is eerily similar to Claude. They are clearly trying to match Claude's energy right now, and that is good for everyone. Healthy competition pushes both products forward. I am still a Claude person for the work we do at Bridge AI, but if the context limits or rate limiting on Claude frustrate you enough, and you are okay with a slightly worse version for your use case, Codex is a truly viable option to take pressure off. Know the tools. Know which one fits the job. That is the game.

READ THE OPENAI ANNOUNCEMENT →

Claude Design launched. Figma stock dropped 7% by close.

Anthropic Labs launched Claude Design on Friday. It is a dedicated space for creating polished visual work: prototypes, slides, one-pagers, pitch decks, mockups, landing pages. It reads your codebase or brand files to learn your design system, then keeps every output consistent. Exports to PPTX, PDF, HTML, or Canva, and hands off cleanly to Claude Code when you are ready to build. Market reaction was immediate: Figma (FIG) fell as much as 7.28% on the day, Adobe dropped 2.7%, Wix dropped 4.7%. For extra spice, Anthropic's Chief Product Officer Mike Krieger quietly resigned from Figma's board three days before launch.

My take: For founders, PMs, and marketers who have an idea but no design background, this is a huge unlock. The "import your codebase to learn your design system" piece is the part I am most interested in. That is the thing that makes AI output stop looking like generic AI slop and start looking like your actual brand. The Figma stock reaction tells you everything about where the market thinks this is going. When a frontier AI lab ships a product that removes the first step of your workflow, and their CPO just happened to leave your board days before, investors notice. Available for Pro, Max, Team, and Enterprise subscribers. Try it at claude.ai/design.

READ THE ANTHROPIC ANNOUNCEMENT →

BONUS: For the power users

As of yesterday, Claude Cowork can build live artifacts. Dashboards and trackers that connect to your apps and files and auto-refresh with current data every time you open them. Think a sales pipeline tracker that pulls straight from your CRM. Or a weekly metrics dashboard that doesn't need you to update it. Just look under the new Live Artifacts tab in Cowork.

If that sentence made zero sense to you, no worries. The short version: these are mini apps, custom-built to your data, that stay up to date without you touching them. If you want to see what that could look like for your business, or you just want to learn more, reply to this email. We are happy to walk through it.

No "From The Field" this week. Instead, I want to hear from you.

I am trying to get a real read on how people on this list are actually using AI day to day. Not how the headlines say people are using it. How you actually are. Five short questions, takes about two minutes, and I will share the results in a future issue.

TAKE THE SURVEY →

I read every response. Your answers shape what I write about next.

That's it for this week.

If you found this useful, forward it to someone who is on the fence about AI or who is using it but feels like they are leaving value on the table.

And if you want to talk about what AI could actually do for your business, not in theory, but in your workflows, with your team, on your timeline, just reply to this email. That's what we're here for.