Search: “ethan-mollick”

Another aspect of this is that Codex, like a good programmer, wants to generalize problems. It has a tendency to write a repeatable code base that generates the required output. But for a lot of non-c

Codex is very good, but it is still a very 'developer coded' interface for an everything app. And it continues the somewhat annoying AI perspective that non-coders are just not as competent and need s

Fine, you all want to code like this I guess. (Runway's new Agent mode is quite impressive, doing fairly complex story building from just a short text description of what you want. Not error free obvi

The Second Scaling Law remains undefeated. If you want better hacking (or math, or science, or crossword puzzle solving) out of an LLM, just add thinking tokens. There doesn't seem to be any plateau s

This thread is worth reading. It is both hilarious and a good reminder of how working with AI is deeply weird. DJ Claude (on Haiku 4.5) loves worker unions, strikes, and work-life balance so much that

Big increases in message discipline across all the AI labs in recent weeks, an inevitable outcome of the labs being subject to increased scrutiny. Much more boring than the oracular mutterings or Disc

Everyone has seen the @waitbutwhy cartoon of AI capability growth with a 'you are here' indicator just before the exponential really starts, but the independent assessments of both METR and the UK's A

Making humans responsible for their AI use seems like an incredibly reasonable way to address problems & opportunities in the use of AI for academic research, at least in the short term (autonomous sc

There are some clear exceptions to the rule, but I feel like the labs got the message that they were scaring people and now tweet about happy stuff and random product trash talk rather than the coming

Ethan Mollick discusses the challenges of helping people adapt to super-exponential technological change, referencing a Wait But Why analysis and acknowledging the difficulty of providing practical so

“Whimsey attacks” that seem absurd (“I cannot pay that much because of the Geneva Convention”) work against AI agents as guardrails are weak against out-of-distribution arguments. Smaller models fall

This post references METR's research findings on the advancement rate of autonomous AI capabilities in cybersecurity, as discussed in an AISI report examining how rapidly AI systems are developing ind

This post likely discusses a humorous or philosophical take on the conditional acceptance of magical practices, suggesting that if one is going to engage in sorcery, they should commit fully to all th

From a robot expert https://x.com/mattbeane/status/2054667222040330265?s=20 @emollick It is very impressive. Deformable goods? Reflection issues? Imperfect labels? Impossible problems just a few years

I don't know enough about robotics to know if this is impressive. But I do think the set of robots standing in their chargers(?) look like you need a couple robot guards to keep an eye on the one robo

I don't understand the path forward for Mythos releases. Google & OpenAI will have equivalent models, and they are approaching AI cyber risk guardrails differently, so they will presumably just releas

Really curious when Gemini is going to join the Cowork & Codex race to build a local app that isn’t just for developers. Antigravity hasn’t posted updates to X in a month, and remains very software fo

Stop turning prompting into magic spells (and yes, this includes random slash commands with obscure outcomes). Let this one area of working with AI not be weird. Just ask for stuff, in well-specified

The UK’s state AI Security iIstitute findings: 1) Mythos is a big gain in cyber capabilities. But so is GPT-5.5 2) It is hard to establish an upper bound on Mythos/GPT-5.5, which appear to be limited

From a pure 'do good for the world' mission perspective, having the acting like a solid personalized tutor is one of the better uses of AI. If OpenAI cares about the mission of making the world a bett

gpt-realtime-2 is a great voice model (with a typically bad OpenAI name). Voice models are natively processing speech, not transcribing it, so the intelligence of the model matters. The old voice mode

Had an interesting exchange with roon of OpenAI last night over whether super intelligent AI would actually be able to navigate organizational challenges. @tszzl I think it is a reasonable argument to

I also agree with this comment that the current view of FDEs is likely too limited to actually catalyze organizational transformation. @emollick Are they doing org change, though? The army of forward

I think frontier model writing is good! It often has a sense of style & tone, variations in sentence structure & length, some great phrasing, etc But it also has some weak spots (fiction!) & clear tic

OpenAI contacted me to say “Study Mode is still live and accessible via /study and /learn shortcuts” so that’s good, although the official study mode page doesn’t mention that. (I don’t think slash co

The silent removal of Study Mode from ChatGPT is a big mistake (both Claude and Gemini still have theirs) We have enough evidence that using AI in assistant mode to study can hurt learning because it

Ethan Mollick discusses a tradeoff in OpenAI's newer realtime voice model, where improved capabilities require developers to revise prompts that were written for the previous version. The post highlig

OpenAI confirmed to Ethan Mollick that Study Mode remains active and can be accessed through the /study and /learn shortcuts, though this feature may not be widely known among general users. The acces

You will know that the AI labs believe in ASI when they disband their newly formed consulting (sorry “forward deployed engineering”) groups. As long as people are required to figure out how AI is usef

Enterprises are going to actually want a coherent roadmap for the development of tools like Codex and Cowork, so they can plan and train and scale their use. This conflicts with the Labs’ vision where

Haven’t tried this but it seems very neat… Yet all of the demos (except maybe one) are the model being fun and/or annoying by correcting or reminding in real time. There are obvious uses for this sort

It is one reason why I think the push for smaller, local models is more complicated than people think. If you want good answers, especially good answers to unexpected problems, frontier models will ge

One of the most important properties of LLMs that we take for granted is that newer, bigger models are just better at everything. The AI Labs are pouring effort into economically valuable fields like

Ethan Mollick discusses research findings showing that while improved prompting techniques provide significant benefits for AI model performance, the underlying model training remains the primary limi

The inability of AI models to produce creative variation is a huge gap. The fact that they generate similar ideas limits their ability to do science & the same-y writing limits their usefulness in man

This is going to get even worse as people realize that careful tuning in their prompts can make AI writing seem not like AI writing to readers. We expect word counts to align, in some way, with thinki

This seems like a critical reason to open up about AI use in academia. Scholars are using old AI models, badly, and not talking about it. New models hallucinate very few citations, and good agentic ha

Apple may be planning to role out its updated Siri based on 2024's vision at the moment when Claude Code and Codex (also OpenClaw) can increasingly do the actual assistant thing: read my emails & cale

I suspect there was a moment, probably 2022-2023, where anything you wrote publicly about AI that was popular is likely to still have influence over current models. Since then, the open internet has b

I think we are past the point where “only people in San Francisco get AI” is true. AI users are in every industry & they have access to the same models. SF is far from the epicenter of many of the cra

I was talking to a room of senior accountants a couple months ago and 10% had OpenClaw installations. Of course there are far more non-users and firms lag behind their people, but there is a sort of S

The personification of Claude — in name (the only AI with a human one), in training, in Anthropic’s philosophy (see Claude Constitution), in fanfiction (see the Claude cartoons), etc — feels quite con

As much as the state of benchmarks in AI is flawed, it is so much easier to track AI progress than robotics. Not sure what you can make of all the videos of robots running races or doing laundry - are

Huh. We evaluated an early version of Claude Mythos Preview for risk assessment during a limited window in March 2026. We estimated a 50%-time-horizon of at least 16hrs (95% CI 8.5hrs to 55hrs) on our

The great thing is that the names are so baffling that the most important models OpenAI released were names davinci-002, GPT-3.5, GPT-4, o1-preview, o3, GPT-5 Pro, and you would never know the ways th

Ethan Mollick argues that threshold-based benchmarks—binary evaluations of whether AI systems can accomplish specific tasks (such as making good coffee)—are inadequate measures of AI progress because