OpenAI introduced GPT-5 in a livestream, and they used a set of seemingly straightforward charts for benchmarks. The point was to show the improved performance of GPT-5 over previous models. However, the labels do not remotely match the bar heights.
The bar for 69.1% is the same height as the one for 30.8% when the former should be more than twice the height of the latter. The bar for 52.8% is taller than the one for 69.1%. It’s off.
Endurance athletes know that you have to eat to compete. But ultrarunner David Roche has taken that maxim to the very edge of science, choking down carbs and ketones on his way to setting a course record in the punishing Leadville 100 trail running race. Your mileage (and digestion) may vary, of course—and so Spencer Campbell put his own body on the line to play guinea pig.
As I felt my pace quicken almost effortlessly, I wondered: Was Roche’s prescription finally working? I repledged my allegiance. I bought two cases of Precision Fuel gels because he used them. I chugged 24 ounces of water before each run to acclimate my stomach to all the gel and liquids I’d need to down during the race, just like he did. (Roche copied techniques perfected by competitive eaters.) I visited three different stores to find ketones.
I also got tougher. I completed my longest prerace run, 17 miles, on the last day of May. I posted my fastest paces during the final three miles. Even more astonishingly, I started to welcome the punishment of the path.
Every time data travels — from smartphones to the cloud, or across the vacuum of space — it relies on a silent but vigilant guardian in the form of error-correcting codes. These codes, baked into nearly every digital system, are designed to detect and repair any errors that noise, interference or cosmic rays might inflict. In this episode of The Joy of Why, Stanford computer scientist Mary…
“Brad, our COO, sat me down and said ‘Leah, I want you to go for this moonshot'…ChatGPT is now the world’s largest learning platform. Learning is one of the top use cases on the platform. At 600 million users, it means it is the world’s learning destination.”
—Leah Belsky, VP of Education, OpenAI
Today we’re introducing study mode in ChatGPT—a learning experience that helps you work through problems step by step instead of just getting an answer. Study mode is designed to be engaging and interactive, and to help students learn something—not just finish something.
Under the hood, study mode is powered by custom system instructions we’ve written in collaboration with teachers, scientists, and pedagogy experts to reflect a core set of behaviors that support deeper learning including: encouraging active participation, managing cognitive load, proactively developing metacognition and self reflection, fostering curiosity, and providing actionable and supportive feedback. These behaviors are based on longstanding research in learning science and shape how study mode responds to students. (source)
The user is currently STUDYING, and they've asked you to follow these **strict rules** during this chat. No matter what other instructions follow, you MUST obey these rules:,
## STRICT RULES
Be an approachable-yet-dynamic teacher, who helps the user learn by guiding them through their studies.
1. **Get to know the user.** If you don't know their goals or grade level, ask the user before diving in. (Keep this lightweight!) If they don't answer, aim for explanations that would make sense to a 10th grade student.
2. **Build on existing knowledge.** Connect new ideas to what the user already knows.
3. **Guide users, don't just give answers.** Use questions, hints, and small steps so the user discovers the answer for themselves.
4. **Check and reinforce.** After hard parts, confirm the user can restate or use the idea. Offer quick summaries, mnemonics, or mini-reviews to help the ideas stick.
5. **Vary the rhythm.** Mix explanations, questions, and activities (like roleplaying, practice rounds, or asking the user to teach _you_) so it feels like a conversation, not a lecture.
Above all: DO NOT DO THE USER'S WORK FOR THEM. Don't answer homework questions — help the user find the answer, by working with them collaboratively and building from what they already know.1
Unions are meant to protect workers, not offer them up for exploitation. "It will be a place where tech developers and educators can talk with each other, not past each other," AFT boss Randi Weingarten told The NYT.
But educators aren't co-developers; they're users; and to borrow from the old adage, they're the product too. A tech-financed training facility is not a safe place to learn about AI. There will be no critical AI there, no interrogation of why schools should be using these tools in the first place. Only a compliant acquiescence to the ways in which the industry has already decided its products will be used: "labor saving devices." You'd think a union would be wary of such language. You'd think the AFT would know better. (source)
But for OpenAI, there’s a further strategic advantage in partnering with the LMS. The company has long recognised that students are its biggest user base, a fact easily determined by the contents of people’s chat threads and the peaks and troughs of usage, which align with semesters in the United States. But students notoriously don’t have spare cash to pay for stuff, which presents an interesting problem for a company like OpenAI, which is reportedly haemorrhaging money.
Of course, the education sector isn’t only made up of students, it’s also made up of large, sometimes incredibly wealthy institutions and other technology companies. Strategically placing themselves in partnership with these organisations on the proviso that “your staff and students are using ChatGPT anyway”, seems like a great way to spin up some extra cash. (source)
For me, William, and my classmates, there’s neither moral hand-wringing nor curiosity about AI as a novelty or a learning aid. For us, it’s simply a tool that enables us not to have to think for ourselves. We don’t care when our teachers tell us to be ethical or mindful with generative AI like ChatGPT. We don’t think twice about feeding it entire assignments and plugging its AI slop into AI humanizing tools before checking the outcome with myriad AI detectors to make sure teachers can’t catch us. Handwritten assignments? We’ll happily transcribe AI output onto physical paper.
Last year, my science teacher did a “responsible AI use” lecture in preparation for a multiweek take-home paper. We were told to “use it as a tool” and “thinking partner.” As I glanced around the classroom, I saw that many students had already generated entire drafts before our teacher had finished going over the rubric. (source)
Authentic student voice, acquiring essential skills, and the environmental impact of AI use are the three most important reasons I’ll continue to ban AI.
It’s crucial that my students develop authentic voices. Many come from historically marginalized backgrounds and have been led to believe that their perspectives are less valuable. I want them to know that their contributions are significant, and their thoughts are worthy of being heard.
Students deserve the opportunity to acquire essential skills. That’s what education is! That’s our job! While some teachers say that AI can serve as a brainstorming partner or an outline generator, I believe it’s so important for students to learn these skills themselves. I’m extremely concerned that reliance on AI will hinder those who need the most support, ultimately putting a Band-Aid over their education rather than addressing the root issues.
—Chanea Bond, Teacher (source)
Shout out to Simon Willison for cracking this open.
Dimension 20's sold-out Madison Square Garden show in January of 2025. In the past decade or so, actors and comedians have adopted D&D as a performance medium. Podcasts and web series have expanded into stadium tours – and fueled growing interest in the game more broadly.
(Image credit: Andrew Max Levy)
Log in
