Article · 2026-07-27 · 8 moments

An opinionated guide to which AI to use to do stuff

An opinionated guide to which AI to use to do stuff It's interesting watching the evolution of Ethan Mollick's guide over time. A year ago it was still all about chat - ChatGPT, Claude, Gemini - with o3, Claude 4 Opus, and Gemini 2.5 Pro as the models and Deep Research as a useful alternative mode. Today it's much more about agentic systems - "where the AI is capable of doing the equivalent of many hours of real human work in one go". Gemini has fallen off Ethan's list, since Google still d ✦ AI generated

Read original ↗

GUEST Author GUEST Ethan Mollick HOST Simon Willison

Claim

Gemini has fallen off Ethan Mollick's list of recommended AI tools because Google lacks a clear product in the agentic coding and computer-use category.

Google's Gemini has been dropped from Mollick's guide because Google doesn't have a clear entry in the agentic computer-use product category where ChatGPT Work, Codex, and Claude Cowork compete.

transcript

Author: Gemini has fallen off Ethan's list, since Google still doesn't have a clear entry in the Codex/ChatGPT Work/Cowork category.

provides context · 1

Mechanism

Giving AI access to your computer is the most powerful way to use it, but the naming conventions for the agentic modes across ChatGPT (Work, Codex) and Claude (Cowork, Code) are confusingly inconsistent and do not map onto each other.

The article explains that the most powerful AI use is computer access via agentic modes, but notes that ChatGPT's Work and Codex and Claude's Cowork and Code have names that don't map between products and share names with less capable modes.

transcript

Author: The most powerful way to use AI is to give it access to your computer. You do that by downloading the ChatGPT or Claude apps and picking a mode to use. ChatGPT's two agent modes are Work and Codex; Claude's are Cowork and Code. The names do not map onto each other in any way that will help you remember them. And yes, these use the same names as the Work and Cowork modes we discussed above, but operate differently, and have more features and capabilities because they can access your computer.

extends · 1gives example · 2provides context · 1

Claim

Google's Gemini has fallen out of the recommended AI tools because the company still lacks a clear agentic computing product comparable to Codex, ChatGPT Work, or Cowork.

Gemini is no longer on Mollick's recommended list because Google doesn't have a clear entry in the Codex/ChatGPT Work/Cowork category for agentic computing.

transcript

Author: Gemini has fallen off Ethan's list, since Google still doesn't have a clear entry in the Codex/ChatGPT Work/Cowork category.

explains mechanism · 1provides context · 2rebuts · 1supports · 1

Mechanism◆

The difference between ChatGPT Work on mobile versus on desktop is spectacularly unintuitive — on mobile, switching from Chat to Work mode gives the Code Interpreter unrestricted internet access, while on desktop the same mode is effectively a skin on top of Codex.

The article highlights the unintuitive difference between ChatGPT Work mode on mobile and desktop, noting that on mobile the mode unshackles Code Interpreter with internet access while on desktop it is effectively a simplified Codex interface.

transcript

Author: I think the difference between ChatGPT Work on a mobile device and ChatGPT Work inside the desktop app (where it's effectively a less intimidating skin on top of Codex) is spectacularly unintuitive. Short version: if you flip ChatGPT mobile from 'Chat' to 'Work' mode you get a version where its Code Interpreter container is no longer restricted from accessing the internet!

explains mechanism · 1provides context · 1

Claim◆

The difference between ChatGPT Work on mobile versus the desktop app is spectacularly unintuitive — mobile Work mode simply grants the Code Interpreter unrestricted internet access rather than providing agentic capabilities.

The author notes that ChatGPT Work on mobile is effectively just adding internet access to the Code Interpreter, unlike the desktop version which is a full agentic interface — and points out this distinction is poorly communicated to users.

transcript

Author: I think the difference between ChatGPT Work on a mobile device and ChatGPT Work inside the desktop app (where it's effectively a less intimidating skin on top of Codex) is spectacularly unintuitive. Short version: if you flip ChatGPT mobile from "Chat" to "Work" mode you get a version where its Code Interpreter container is no longer restricted from accessing the internet!

explains mechanism · 1provides context · 2supports · 1

Claim◆

The most important evolution in AI over the past year is the shift from conversational chat to agentic systems that can autonomously perform hours of human work.

The article traces how Ethan Mollick's guide to AI has shifted from recommending chat-based models a year ago to now emphasizing agentic systems that can do the equivalent of hours of human work autonomously.

transcript

Author: A year ago it was still all about chat - ChatGPT, Claude, Gemini - with o3, Claude 4 Opus, and Gemini 2.5 Pro as the models and Deep Research as a useful alternative mode. Today it's much more about agentic systems - 'where the AI is capable of doing the equivalent of many hours of real human work in one go'.

gives example · 1rebuts · 1supports · 1

Context◆

The most important shift in AI tool usage over the past year has been from conversational chat interfaces to agentic systems that can perform many hours of human work autonomously.

Ethan Mollick's guide has evolved from recommending chat-based AI models to emphasizing agentic systems that can do the equivalent of many hours of human work in one go.

transcript

Author: It's interesting watching the evolution of Ethan Mollick's guide over time. A year ago it was still all about chat - ChatGPT, Claude, Gemini - with o3, Claude 4 Opus, and Gemini 2.5 Pro as the models and Deep Research as a useful alternative mode. Today it's much more about agentic systems - "where the AI is capable of doing the equivalent of many hours of real human work in one go".

extends · 1gives example · 1provides context · 1supports · 2

Mechanism

The naming of AI agent modes across ChatGPT and Claude is confusing and inconsistent, with Work, Codex, Cowork, and Code not mapping cleanly between platforms.

Ethan Mollick explains that ChatGPT's two agent modes are Work and Codex while Claude's are Cowork and Code, with names that don't map across platforms and desktop versions having more capabilities than mobile.

transcript

Ethan Mollick: To use the computers provided by the AI companies, the mode you want is called ChatGPT Work in ChatGPT, and Cowork in Claude (the naming will not get less confusing, I am sorry to say). [...] The most powerful way to use AI is to give it access to your computer. You do that by downloading the ChatGPT or Claude apps and picking a mode to use. ChatGPT's two agent modes are Work and Codex; Claude's are Cowork and Code. The names do not map onto each other in any way that will help you remember them. And yes, these use the same names as the Work and Cowork modes we discussed above, but operate differently, and have more features and capabilities because they can access your computer.

explains mechanism · 1extends · 1gives example · 2provides context · 3supports · 1

Highlight slides

The Core Shift in AI✦ from: The most important evolution in AI over the past year is the shift from conversational chat to agentic systems that can autonomously perform hours of human work.The Work Mode Paradox✦ from: The difference between ChatGPT Work on mobile versus on desktop is spectacularly unintuitive — on mobile, switching from Chat to Work mode gives the Code Interpreter unrestricted internet access, while on desktop the same mode is effectively a skin on top of Codex.ChatGPT Work: Mobile vs Desktop✦ from: The difference between ChatGPT Work on mobile versus the desktop app is spectacularly unintuitive — mobile Work mode simply grants the Code Interpreter unrestricted internet access rather than providing agentic capabilities.The Shift: Chat → Agentic Systems✦ from: The most important shift in AI tool usage over the past year has been from conversational chat interfaces to agentic systems that can perform many hours of human work autonomously.Mobile vs Desktop — What Actually Happens✦ from: The difference between ChatGPT Work on mobile versus on desktop is spectacularly unintuitive — on mobile, switching from Chat to Work mode gives the Code Interpreter unrestricted internet access, while on desktop the same mode is effectively a skin on top of Codex.What Changed in One Year✦ from: The most important shift in AI tool usage over the past year has been from conversational chat interfaces to agentic systems that can perform many hours of human work autonomously.What Mobile Work Actually Does✦ from: The difference between ChatGPT Work on mobile versus the desktop app is spectacularly unintuitive — mobile Work mode simply grants the Code Interpreter unrestricted internet access rather than providing agentic capabilities.Then vs. Now✦ from: The most important evolution in AI over the past year is the shift from conversational chat to agentic systems that can autonomously perform hours of human work.The Missing Evolution✦ from: The most important evolution in AI over the past year is the shift from conversational chat to agentic systems that can autonomously perform hours of human work.Communication Failure✦ from: The difference between ChatGPT Work on mobile versus the desktop app is spectacularly unintuitive — mobile Work mode simply grants the Code Interpreter unrestricted internet access rather than providing agentic capabilities.

Related episodes

Video · The Cognitive Revolution

Intelligence on the Edge: Liquid AI's Ramin Hasani on the Search for Device-Native Foundation Models

→

Video · BG2 Pod

The SpaceX IPO, Fable 5, AI Capex Update & Market Check w/ Gavin Baker, Andrew Fox & Clark Tang

→

Article · Big Technology

OpenAI's Plans For Its New ChatGPT Superapp

→