OpenAI & Anthropic are targeting something WAY bigger

MMaximilian Schwarzmüller
Computing/SoftwareBusiness NewsInternet Technology

Transcript

00:00:00for these ai companies anthropic open ai and so on coding is solved now again for them i don't
00:00:07share that notion for me ai does way too many mistakes and introduces way too much bloat and
00:00:15bugs and errors into the code base for me to blindly trust it but yeah who knows how things
00:00:20will change what's interesting of course is that for these ai companies code is kind of only the
00:00:27first step and that's really important to understand obviously we developers we are the
00:00:34pioneers we're amongst the first that really mass adopt these llms and these agentic coding tools
00:00:42obviously because well they're agentic coding tools right there were meant for coding so obviously
00:00:47they're for us and we're using those we're using those and you probably are using codex cloud code
00:00:52github copilot cursor any of these to help you work on your code now what's interesting and what's
00:00:59important to understand is that for these companies code is just the first step they want you to use
00:01:07codex or cloud code for almost everything this is an official post it is about more than coding and indeed
00:01:16in codex you can also generate images you can spin up a chat session ask it to generate an image and it
00:01:23will it gives you a special skill for that it should also activate that on its own of course and it will
00:01:29use the gpt image 2 model provided by openai to generate an image and you can then also follow up on
00:01:35that include a logo and stuff you can do that in the codex app you can also analyze sales in there or
00:01:43anything you can do data analytics there so stuff that has nothing to do with code why because of
00:01:48course many things that you do on a computer can be solved with code and with the command line without
00:01:55you even knowing that code is involved of course you as a developer do but keep in mind we're a relatively
00:02:01small group for openai and anthropic the more interesting part of the pie is all the rest all
00:02:07the other white collar work all the people that spend their days working in excel sheets or word
00:02:13documents or crunching numbers working with pdf documents that is the more interesting part we are
00:02:19just small a small number of early adopters we the developers but it's the the other part that matters
00:02:26even more so naturally these companies want to turn their apps into super apps that can do it all and
00:02:32therefore of course fittingly openai announced computer use which is not available in the eu yet
00:02:38so i couldn't try it yet but which is all about codex being able to control any application on your
00:02:45machine so that it can even work with applications that are hard to control with just code and the command
00:02:52the command line now one important note and shameless plug here i created courses on cloud code
00:02:58and codex and my colleague manual created a course on cloud co-work to which i'll get back in a second
00:03:04and whilst of course the cloud code and codex courses are all about using these tools for
00:03:10coding we have the co-work course for teaching you the cloud co-work tool that is about doing non-coding
00:03:17tasks on your computer with ai and in the codex course i have all the content about coding but
00:03:23also some content on performing non-coding tasks with codex so if you want to dive in deep and learn
00:03:30all the nitty-gritty details and tricks and tips about cloud code codex and co-work these courses may be
00:03:36interesting to you and speaking of co-work anthropic is following a different strategy they also have cloud code
00:03:44and obviously they have the cloud code cli which is probably the most common way for developers to work
00:03:51with cloud code but they also have their desktop app and unlike open ai they don't have the one big
00:03:56super app which clearly is the goal of open ai to turn codex into that universal super app that can
00:04:03do it all and that works for both coders and non-coders anthropic at least right now has that split
00:04:09you have the code part in the desktop app here where you can interact with cloud code as you can from
00:04:16the command line essentially use your skills ask it to do stuff and so on but then you also got a chat
00:04:22area where you can guess what chat with cloud just as you can do in the web interface and you got the
00:04:28co-work area and the co-work area is essentially like cloud code but optimized for non-coding tasks in codex you
00:04:37also have a setting where you can switch to a different work mode to kind of tweak some settings
00:04:43for non-techy people so clearly that is their goal capture the overall market of white collar work
00:04:53anthropic with the model of having different parts in that app open ai with the model of having the one super
00:05:01app and of course we will very likely also see other applications come up third-party open source
00:05:07applications like the amazing py agent here for example which is primarily about coding right now
00:05:13but which again you can use for anything because ultimately as i mentioned with code you can do a lot
00:05:19on your machine so you can use agent like this too to ask it to do something in an excel sheet or a pdf
00:05:24document and it will figure out a way of doing so with code and command line tools and that's therefore
00:05:31their vision now the interesting part of course is that for me coding isn't solved yet not even close
00:05:39and we'll see how that changes and how we as developers evolve and clearly our role is evolving and
00:05:45clearly we must use these tools and learn how to use them efficiently but i guess we're still all figuring out
00:05:51how to do that and i would assume the same happens for other areas of work these ai tools can obviously
00:06:00assist with lots of work the danger of course always is that you as a human get pretty lazy and you might
00:06:08not care anymore about the excel sheet that was produced for example you might not double check it and
00:06:14we'll see a similar learning process there as we see it for us developers and we'll also see if that one
00:06:22super app strategy works out for open ai or if it turns out to be a bit too confusing for normies to
00:06:30have all the coding stuff in there too which is not super invisible even if you do switch to that
00:06:37non-coding specific view but yeah their vision is clearly that code is just the beginning and we'll see
00:06:46whether that turns out to be true or not

Key Takeaway

OpenAI and Anthropic are evolving their AI coding tools into universal productivity platforms designed to automate all white-collar computer work, not just software development.

Highlights

  • OpenAI and Anthropic are expanding beyond code generation to dominate general white-collar computer tasks.

  • OpenAI is pursuing a 'super app' strategy by consolidating coding, data analytics, and image generation within the Cursor application.

  • Anthropic separates its desktop application into distinct workspaces, including a dedicated 'Claude Work' area optimized for non-coding tasks.

  • The 'Computer Use' feature allows AI to control desktop applications directly, bypassing limitations of traditional command-line interfaces.

  • Developers act as early adopters, but the primary long-term market for these AI companies is the broader white-collar workforce using Excel, Word, and PDF documents.

Timeline

Shift from Coding to General Productivity

  • Coding serves as the initial adoption vector for AI agentic tools.
  • AI companies prioritize capturing the broader white-collar market over the small developer demographic.
  • Tools like Cursor enable non-coding tasks, such as data analytics and image generation, directly within the development environment.

While developers are the primary early adopters of agentic coding tools, these technologies are designed for a much larger user base. Companies like OpenAI and Anthropic treat coding as the first step in a broader strategy to automate tasks performed in applications like Excel, Word, or PDF readers. By integrating features like data analysis and image generation into coding tools, these platforms aim to become all-encompassing productivity solutions for general office work.

Strategy Divergence: Super Apps vs. Modular Workspaces

  • OpenAI integrates 'Computer Use' to allow AI to control any desktop application.
  • OpenAI utilizes a 'super app' model that combines all functionalities into one interface.
  • Anthropic maintains a modular app structure, separating coding, chat, and non-coding tasks into distinct workspaces.

OpenAI and Anthropic exhibit different approaches to user interface design as they expand their reach. OpenAI is building a unified 'super app' by centralizing diverse AI capabilities, including new 'Computer Use' features that enable the model to interact with any application on a user's machine. Conversely, Anthropic segments its desktop app, offering dedicated sections for coding and non-coding workflows to provide a structured experience for different types of users.

Future Risks and Market Evolution

  • Third-party, open-source agents like 'PyAgent' are emerging to perform general computer automation.
  • Human oversight remains necessary to prevent errors in AI-generated work, particularly in documents like Excel sheets.
  • The effectiveness of the 'super app' strategy remains unproven for non-technical users.

The market for AI-driven automation is expanding through both centralized commercial platforms and third-party, open-source projects. As these tools become capable of executing complex workflows, human users face the risk of becoming overly dependent and neglecting essential verification steps. The ultimate success of these models depends on whether a unified 'super app' design is intuitive enough for general users or if a more specialized approach proves more effective.

Community Posts

No posts yet. Be the first to write about this video!

Write about this video