- AI Sidequest: How-To Tips and News
- Posts
- A $300K editing job? Yowza.
A $300K editing job? Yowza.
Plus, a big AI-fiction study, a Sol review, and a huge leap in open-source AI.
Mignon Fogarty
July 20, 2026
In partnership with
Issue 112
On today’s quest:
— Get a load of these salaries!
— Or maybe AI fiction is bad?
— My top productivity tool: Wispr Flow
— Fable-level open-source models have no guardrails
— People seem to like GPT-5.6 Sol
— Self-driving cars could ease traffic
Get a load of these salaries!
Anthropic has a job opening for a standards editor that pays nearly $300K per year (requiring just 25% of time in the office). And the wild thing is that this doesn’t even seem to be a super high-level position: the description includes “you’ll … work closely with and learn from experienced writers and editors.”
There are other editing and writing jobs in the same range or higher.
I don’t think I’m dramatically out of touch on editing salaries when I say, “This seems unusually high.” My strong guess is that it would also include pre-IPO stock options. Yowza.
Or maybe AI fiction is bad?
In the last newsletter, I told you how people keep accidentally liking AI fiction because AI-generated stories seem to keep winning awards and being chosen for anthologies. But this week, 404 Media has a story titled AI Fiction Is Easy to Detect Because It's Stupid and Bad, Research Finds (subheading “ChatGPT uses too many dream sequences and Gemini won’t stop describing characters”).
The impressively scaled study from Google DeepMind and researchers at the University of Maryland, College Park, looked at 50,000 AI-generated short stories and found that “AI stories over-explain themes and favor tidy, single-track plots while human stories frame protagonists’ choices as more morally ambiguous and have increased temporal complexity,” as well as further narrative differences.
So which is it? Are AI-generated short stories good or bad? I guess it depends on what you like and how many you look at. Even if they are bad en masse, a subset could be what we consider traditionally “good,” and I also imagine at least some people who submit AI-generated stories to contests and anthologies don’t send in the one-shot output of a prompt like what was used in the study.
My top productivity tool: Wispr Flow
Ignore Wispr Flow’s ad copy below and click the link at the bottom to try it free. I save at least 15 minutes a day by talking to my computer instead of typing (and I’m a fast typist). I recommend Wispr Flow to everyone I know in real life, and I think you’ll like it too.
Your prompts are leaving out 80% of what you're thinking.
When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning — constraints, edge cases, examples, tone — and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.
89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Try Wispr Flow free — works on Mac, Windows, and iPhone.
Fable-level open-source models have no guardrails
Less than two months ago, the U.S. government forced Anthropic’s Fable model off the market over national security concerns, but this week, a Chinese company has released an open-source model called Kimi-K3 that rivals Fable on AI benchmarks and appears to have no safety guardrails.
The AI Daily Brief podcast discussed the implications at the 24:00 mark, but it’s not great.
A story I saved six weeks ago described someone intentionally building a vulnerable app and then seeing which LLMs could hack it. Commercial models were interesting because some honored their guardrails and refused to hack, while others would have succeeded if they hadn’t hit their budget limits. At the time, the best open-source model only succeeded 30% of the time, but I bet Kimi-K3 would do much better today. Things are changing fast.
Further, demand for K3 is so high that the company had to suspend new subscriptions.
Axios reports that the U.S. government is considering limiting access to Chinese models within the U.S., which seems counterintuitive since that would leave them in the hands of people who might want to attack the U.S., but keep them out of the hands of U.S. companies who would want them for defense.
People seem to like GPT-5.6 Sol
The top model from OpenAI, GPT-5.6 Sol, was new when I sent the last newsletter, but reviews are now looking good. I always like Every’s comprehensive reviews, and they say Sol is their favorite model for day-to-day collaboration (although they still prefer Fable for the toughest knowledge work).
Their analysis has an interesting section about writing too, saying, “Opus uses plainer language inside more complicated syntax; Sol produces a page that is easy to scan and lines that take more work to understand.” They said Sol was the worst writer out of the box but was better than Opus when given material and context to work with. The whole long review is worth reading if you want to understand when it makes sense to use Sol.
Self-driving cars could ease traffic
I recently got back from a long road trip, which involved sitting in multiple traffic jams, so I was especially fascinated to read about a 2022 study that found mixing self-driving cars into normal traffic diminished “phantom traffic jams” that seem to be caused for no reason. The paper built on earlier studies done on circular tracks, which found that these mystery traffic jams are caused by the human tendency to brake and speed up slightly erratically, causing waves of backups to travel through the line of cars.
For fun
I haven’t tested this … yet. :)
Quick Hits
My favorite recent pieces
Publishers Are Preparing to Opt Out of Google Search [The hosting company Cloudflare is making opting out of Google indexing the default state for new sites to combat AI scrapers, and big publishers are considering the benefits of the once unthinkable — not being in Google search results since traffic from Google is plummeting anyway. However, in the long run, this could be a disaster for the information environment, which is already being flooded with low-quality information.] — AdWeek
What Would You Do if Labor Was Free? [A utopian perspective, for sure, but it includes a lot of things that could happen with cheap robots that I wouldn’t have considered] — It Can Think
Modern-day oracles or bullshit machines? [I disagree with a few of their final conclusions, but I still think the world would be much better if everyone read their overview of what LLMs can and can’t do. It’s an exceptional educational piece. (h/t Katharine O’Moore-Klopf)] — Carl T. Bergstrom & Jevin D. West
It’s Not the AI. It’s the Lawyer: We Don’t Have a Hallucination Problem, We Have a Serious Ethics Problem [People were stupid and unethical long before AI. Don’t treat AI-related lapses as something special.] — AI Law Librarians
Using AI
5 AI Failures That Aren't Going Away — Why Try AI
Agents
Bad stuff
Meta Used AI to Target Workers With Medical Conditions for Layoffs, Lawsuit Claims — US News & World Report
The business of AI
IBM stock craters after company issues second-quarter earnings warning [The AI infrastructure build-out is causing customers to shift spending.] — CNBC
Companions
China cracks down on AI companions, forcing millions to break up with virtual partners [China deleted millions of online companions without notice causing significant distress for users.] — ABC News
OpenAI's First AI Device Will Be a Portable Smart Speaker [They’re specifically referring to it as a companion, saying it will “connect on a humanlike level with users."] — MacRumors
Climate & Energy
The Whole Big Data Center Fight Is Not Really About AI [interesting history]— Manual Transmission
Education
Fine, Let’s Talk About AI Bans [A reaction to the policy outlined in the link above.] — RIPS Law Librarian Blog
Despite the growth of some AI schools like Alpha, research doesn’t show that AI tutors are better than human teachers — The Conversation
Claude for K-12 Teachers [Anthropic now provides free premium Claude for teachers] —Anthropic
Government
I’m laughing
Anthropic's newest ad is creeping people out [This ad is so tone deaf it’s practically an anti-AI ad. Pat broke out in astonished laughter when I showed it to him, and Sam Altman posted that he originally thought it had to be fake. It’s not funny, but it’s funny that they released it.] — TechCrunch
Job market
A New Look at AI’s Impact on Jobs [“Firms making the largest AI investments grow employment by roughly 10% following adoption, while low-intensity adopters see no statistically significant change. Entry-level headcount rises 12% for high-intensity adopters. Gains emerge gradually and are broad across roles, including engineering, sales, administration, and customer service.”] — Ramp
Managers, and their belief in the value of AI & how to use it, determines the value they actually get from AI — SSRN (via Ethan Mollick)
Philosophy
If We Must Act Now on AI – What Actually Should We Do? [Pondering “what gives people agency, what creates dependence, and what humans do when necessity no longer tells them who to be.”] — Turing Post AI
A Jewish Case for AI work exemptions [or why should Chicago Pope have all the fun?] — Life Is a Sacred Text
A Framework for Frontier AI and the Dawning of a New Age — Demis Hassabis
Once again we are told AI may be conscious – I study consciousness, and I have my doubts — The Guardian
Publishing
U.S. Fiction Audiobook Consumers Rate Spoken Multi-Cast Higher Than Human Narration [with caveats from the study design] — SSRS
Someone Used A.I. to Write an Unauthorized Biography of Me. I Don’t Recommend Reading It. — The New York Times
Robotics
Boston Dynamics touts robot dogs to fix the ‘porch gap’ — Boston Dynamics
Floating Companion: Exploring Design Space for Soft Floating Robots in Indoor Environments [floating robot whale!] — YouTube
Science & Medicine
Security
Head of Federal AI Safety Org Resigns After only 3 Months — Daily Signal
Video
Other
How we learned to trust our AI code reviewer at DoorDash [Pairing two models — the open-source Kimi K2.6 with Claude Fable 5 — caught about two-thirds of the known bugs.] — DoorDash
How People Are Really Using AI in 2026 [Includes especially interesting charts about changes in use over time.] — Harvard Business Review
AI can predict how you’ll respond to a survey. But that’s not the same as understanding you — The Conversation
How OpenAI’s Sol Finally Learned Design Taste — Design Arena
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