When AI makes 'janky' feel right

Plus: AI fog, author scam warnings, Codex in Chrome, and a publishing conference.

Author

Mignon Fogarty
May 11, 2026

Issue 108

On today’s quest:

— The human-forward vibe
— Word watch: AI fog
— Don’t spread your AI experiments too thin
— More than 80% of financial services firms adopting AI
— More on scams targeting authors
— Codex can now do things in your browser
— Free AI in Publishing conference

The human-forward vibe

The New Yorker highlighted a growing trend toward a more amateurish hand-made vibe in design circles to emphasize humanity as a direct backlash against AI art — even for AI companies. For example, the design firm that did the logo for the AI note-taking company Granola said the design is “janky in places, by design, [providing] a contrast with OpenAI’s branding, which is ‘very precise’” and “could be the identity for a defense firm.”

I’ve noticed this in my own tendencies too. I was an art major for a semester (before failing drawing after getting a sinus infection from doing charcoal drawing in a poorly ventilated classroom), and I’ve been wanting to draw again lately even though everything turns out terrible. (h/t Fritinancy)

Word watch: AI fog

The Harvard Business Review highlighted “AI fog” — the inability to plan for the future like you used to because of the uncertainty created by AI. “Given all the things that might change because of AI, it feels like a fog has descended that occludes our ability to see the future,” said Tobe E. Stuart, Director of the Berkeley Entrepreneurship program.

AI fog is especially hard for students who are making expensive long-term commitments like choosing a field of study or deciding to go to graduate school, but it also can make companies hesitant to hire and invest in long-term projects or equipment.

Stuart says individuals should “cultivate psychological and professional agility … the ability to abandon a plan despite having invested in it, to reskill often, to refocus on opportunities as they arise, and to let go of our professional identities so we can adapt them as circumstances evolve.”

Don’t spread your AI experiments too thin

Katie Robberts had great advice in the Trust Insights newsletter: in corporations, one focused AI test project is better than five small test projects.

She argues that trying too many things at once means people can’t truly rally around a project, nobody has the budget they need to take it to the next step, and you can’t gather enough meaningful data to make good decisions.

She recommends picking one project for your test, and it doesn’t have to be the biggest or most impressive. Click through for her advice on how to pick the best project for a test.

More than 80% of financial services firms adopting AI

“More than 80% of financial services firms are adopting AI to some level and 52% are already experimenting with agentic AI, but impact to date is on efficiency gains rather than business model transformation, says a new report by the Cambridge Centre for Alternative Finance.” This FT story is a good overview of the study.

More on scams targeting authors

Since I wrote about the increase in AI scams targeting authors in the last newsletter, I’ve heard from authors who confirm they get these kinds of emails daily, and Macmillan has put out a statement warning authors about scams that use their logos and the names of Macmillan imprints, employees, and other real people in the publishing world. Unfortunately, I also saw someone share one of these emails in a group and get comments that must be real.

AI makes it easy to send emails that seem professional and real, and the scams are complex. They don’t ask for money right away. Please, be wary. If you find an offer interesting, you can do a few things to vet it:

  1. Check the sender’s email address. Often they are made to look similar to a real address, but differ in small ways. For example, Macmillan has a list of domains people are using to impersonate them, including macmillàn.com (with an accented “a”) and macmillanpublishers.com (which sounds credible but isn’t them).

    Copy the domain on the senders email and paste it into your browser to see if it’s real. Often, you’ll just get an error because the site doesn’t exist, and then you know it’s a scam.

    But these days, it’s also trivial to put up a professional-looking website, so finding one doesn’t prove it’s not a scam. If you find a site, you can put it into the Wayback Machine and see how long it’s been online. If it’s new, that’s a good reason to be especially suspicious.

  2. Check the sender’s name. Many scams use generic names that are hard to search, so if your email is from Ashley Smith, that’s another reason to be suspicious (with apologies to all the Ashleys and Smiths out there!). But author scammers are often using the names of real editors and agents, so you have to dig further.

  3. At this point, the scams are so widespread that if everything above checks out and you want to respond, I suggest you still contact the person directly, in some way other than replying to the email. Find them on LinkedIn (and make sure it’s not a new account), call the company, or find the person’s contact information some other credible way. Don’t worry about being weird — if they are really trying to contact you, they’ll think you’re smart for knowing about the scams.

Interestingly, I checked my spam folder to see if I had examples I could show you, and I don’t have any author scams, but I’m getting daily emails that are being sent directly to spam offering to invest in Grammar Girl or buy Grammar Girl that seem like a credible first email a big investor would send. (Good job, Gmail!) I guess the scam databases have me pegged more on the entrepreneur axis than the author axis. These emails were mostly from people with common names from domains that don’t exist when I plug them into a browser.

Codex can now do things in your browser

Codex is OpenAI’s answer to Claude Code, and it now can control Chrome through an extension in what looks like the same way you can with the Claude in Chrome extension. The demo shows it gathering posts from forums to determine sentiment, analyzing them, and making a spreadsheet report; finding receipts in email and filing an expense report through the browser; and more. (Demos)

Doing tasks through direct control of the browser tends to be especially token-intensive, so I can imagine people switching between Claude in Chrome and Codex when they reach their limits.

Free AI in Publishing conference (starts TODAY)

PurePub AI is hosting an online conference about AI in publishing with an extensive line-up of speakers including Jane Friedman and Erin Servais. It’s free to watch live (with a fee for a “watch later” ticket). It starts Monday, May 11, at 8 am Pacific and continues with about one session per day through May 22. (h/t Erin)

Quick Hits

Using AI

A bunch of good ones this week:

How To Force AI to Write More Like You — Christopher Penn, Almost Timely Newsletter

How to use AI to organize your download folder — AI + Education = Simplified

Audio

The business of AI

Anthropic CEO Dario Amodei says company grew 80-fold in first quarter [Anthropic is supplementing its desperate demand for compute by leasing an xAI datacenter that uses especially dirty energy, and by doing so, it was able to immediately increase usage limits.] — CNBC

I’m laughing

Job market

Model & Product updates

Psychology

Robotics

Science & Medicine

Security

Mythos found 271 bugs in Firefox. Previous AI scans found 22. [Nate argues this will change how much people trust human-written code.] —Nate B, Jones

Other

The context window has been shattered: Subquadratic debuts a 12-million-token window [This seems like a big technological breakthrough. Currently, the largest token windows are 1 million.] — The New Stack

What is AI Sidequest?

Are you interested in the intersection of AI with language, writing, and culture? With maybe a little consumer business thrown in? Then you’re in the right place!

I’m Mignon Fogarty: I’ve been writing about language for almost 20 years and was the chair of media entrepreneurship in the School of Journalism at the University of Nevada, Reno. I became interested in AI back in 2022 when articles about large language models started flooding my Google alerts. AI Sidequest is where I write about stories I find interesting. I hope you find them interesting too.

If you loved the newsletter, share your favorite part on social media and tag me so I can engage! [LinkedIn — Facebook — Mastodon]

Written by a human