#Tag
#OpenAI released a standalone #Codex app for #Windows, mirroring the features of the #macOS version. The app allows users to coordinate multiple #codingagents, automate tasks, and utilise a “Skills” section for streamlined workflows. https://www.engadget.com/ai/openai-brings-its-codex-coding-app-to-windows-195345429.html?eicker.news #tech #media #news
#OpenAI released a standalone #Codex app for #Windows, mirroring the features of the #macOS version. The app allows users to coordinate multiple #codingagents, automate tasks, and utilise a “Skills” section for streamlined workflows. https://www.engadget.com/ai/openai-brings-its-codex-coding-app-to-windows-195345429.html?eicker.news #tech #media #news
@jhlagado this is an interesting thread here.
As a quick asside: @ArneBab brings up some counterpoints backed by evidence from Antrhopic to contradict what Yegge claims about the 10x productivity gains of coding agents, but that is not the main point of his article.
The point Yegge is making is that using coding agents is burning everyone out.
I remember a few years ago he wrote a blog post for Sourcegraph on how he used AI coding agents to write code and was amazed at how good it did on a simple task. He said anyone who thought coding agents weren’t good enough were clutching their pearls. Now he is writing about being worried that AI coding agents are like “energy vampires.” He also suggests the burn-out is being caused by achieving results more quickly but that you probably aren’t being paid more for your improved productivity, you are just working more to keep ahead of other people using coding assistants.
What I think it might be is that people aren’t really learning anything new as much as they used to nowadays. Sure, solving programming problems was hard and time consuming, and prevented you from achieving the result for which you get paid. But once you solved the problem, you earn more than just pay, you earn useful experience that you will never forget, which can apply to other projects and make your job easier over all.
If you are doing something that is really manually laborious that you have done a hundred times already and just don’t want spend time doing it again, then sure coding agents are great and useful. But that is what, 10% of a software engineers job? 20% at most maybe? Leaving out administrative work, that other 80% is really about learning and understanding technology. Are we really learning anymore, or are we just letting the machine do our homework for us now, and handing in more and more homework faster?
By the way, @dthompson is one of the people working on the Spritely “Goblins” project, and the Guile “Hoot” compiler which compiles Scheme to WebAssembly.
#tech #software #AI #CodingAgents #GenAI #GenerativeAI #LLMs #ClaudeCode #Anthropic #ChatGPT #SoftwareEngineering
RE: https://toot.cat/@dthompson/116170931695898604
look, no one could have foreseen that the bosses were going to capture all the value of generative AI. we had to ruin the software industry to find out. the solution is a new idea where workers collectively join together to demand fair treatment (needs a name)
@jhlagado this is an interesting thread here.
As a quick asside: @ArneBab brings up some counterpoints backed by evidence from Antrhopic to contradict what Yegge claims about the 10x productivity gains of coding agents, but that is not the main point of his article.
The point Yegge is making is that using coding agents is burning everyone out.
I remember a few years ago he wrote a blog post for Sourcegraph on how he used AI coding agents to write code and was amazed at how good it did on a simple task. He said anyone who thought coding agents weren’t good enough were clutching their pearls. Now he is writing about being worried that AI coding agents are like “energy vampires.” He also suggests the burn-out is being caused by achieving results more quickly but that you probably aren’t being paid more for your improved productivity, you are just working more to keep ahead of other people using coding assistants.
What I think it might be is that people aren’t really learning anything new as much as they used to nowadays. Sure, solving programming problems was hard and time consuming, and prevented you from achieving the result for which you get paid. But once you solved the problem, you earn more than just pay, you earn useful experience that you will never forget, which can apply to other projects and make your job easier over all.
If you are doing something that is really manually laborious that you have done a hundred times already and just don’t want spend time doing it again, then sure coding agents are great and useful. But that is what, 10% of a software engineers job? 20% at most maybe? Leaving out administrative work, that other 80% is really about learning and understanding technology. Are we really learning anymore, or are we just letting the machine do our homework for us now, and handing in more and more homework faster?
By the way, @dthompson is one of the people working on the Spritely “Goblins” project, and the Guile “Hoot” compiler which compiles Scheme to WebAssembly.
#tech #software #AI #CodingAgents #GenAI #GenerativeAI #LLMs #ClaudeCode #Anthropic #ChatGPT #SoftwareEngineering
RE: https://toot.cat/@dthompson/116170931695898604
look, no one could have foreseen that the bosses were going to capture all the value of generative AI. we had to ruin the software industry to find out. the solution is a new idea where workers collectively join together to demand fair treatment (needs a name)
I don’t have FOMO about coding agents or GenAI. If they turn out to be as efficient as the hype says, learning to use a chat window later will be the easy part.
Mastering the craft and knowing what to ask for is the hard part and a better use of my time.
I don’t have FOMO about coding agents or GenAI. If they turn out to be as efficient as the hype says, learning to use a chat window later will be the easy part.
Mastering the craft and knowing what to ask for is the hard part and a better use of my time.
