The way Claude Code is going is exactly what I want out of a agentic coding tool with this "unix toolish" philosophy. I've been using Claude code since the initial public preview release, and have seen the direction over time.
The "golden" end state of coding agents is that you give it a Feature Request (EG Jira ticket), and it gives you a PR to review and give feedback on. Cursor, windsurf, etc, are dead ends in that sense as they are local editors, and can not be in CI.
If you are tooling your codebase for optimal AI usage (Rules, MCP, etc), you should target a technology that can bridge the gap to headless usage. The fact Claude Code can trivially be used as part of automation through the tools means it's now the default way I thinking about coding agents (Codex, the npm package, is the same).
Disclaimer, I focus on helping companies tool their codebases for optimal agent usage, so I might have a bias here to easily configurable tools.
Not sure about that golden end state. Mine would be being in a room surround by screens with AI agents coding, designing, testing, etc. I would be there in the center giving guidance, direction, applying taste, etc…
All conversational, wouldn’t need to touch the keyboard 99% of the time.
I hate using voice for anything. I hate getting voice messages, I hate creating them. I get cold sweats just thinking about having to direct 10 AI Agents via voice. Just give me a keyboard and a bunch of screens, thanks.
If that’s the future, that means a massive reduction in software engineers no? What you are describing would require one technical product manager, not a team of software engineers.
I can easily see this happening in 2-3 years. Some chat apps already have outstanding voice mode, such as GPT-4o. It's just a matter of integrating that voice mode, and getting the understanding and generated code to be /slightly/ better than it is today.
It seems unlikely that any one individual would be able to output a sufficient amount of context for that to not go off the rails really quickly (or just be extremely inefficient as most agents sit idle waiting for verification of their work)
No. The "golden" end state of coding agents is free and open source coding agents running on my machine (or in whatever machine I want). Can you imagine paying for every command you run in your terminal? For every `ls`, `ps`, `kill`? No sense, right? Well, same for LLMs.
I'm not saying "ban propietary LLMs", I'm saying: hackers (the ones that used to read sites like this) should have as their main tools free and open source ones.
> Can you imagine paying for every command you run in your terminal?
Yes, because hardware and electricity aren't free.
I literally DO pay for every command. I just don't get an itemized bill so there's no transparency about it. Instead, I made some lump-sum hardware payment which is amortized over the total usage I get out of it, plus some marginal increase in my monthly electric bill when I use it.
I agree with the sentiment, but isn’t Claude Code (the CLI) FOSS already? (Not sure it’s coupled to Claude the model API either, but if it is I imagine it’s not too hard to fix.)
> Cursor, windsurf, etc, are dead ends in that sense as they are local editors, and can not be in CI.
I was doing this with Cursor and MCPs. Got about a full day of this before I was rate limited and dropped to the slowest, dumbest model. I’ve done it with Claude too and quickly exhaust my rate limits. And the PRs are only “good to go” about 25% of the time, and it’s often faster to just do it right than find out where the AI screwed up.
> The "golden" end state of coding agents is that you give it a Feature Request (EG Jira ticket), and it gives you a PR to review and give feedback on.
I see your point but in the other hand how depressing to be left only with the most soul crushing part of software entering - the Jira ticket.
I personally find figuring out what the product should be is the fun part. There still a need for architecting a plan, but the actual act of writing code isn't what gives me personal joy, it's the building of something new.
I understand the craft of code itself is what some people love though!
Say what you will, but this would have the wonderful side effect of forcing people who write JIRA tickets to actually think through and clearly express what it is they want built.
The moment I am able to outsource work for Jira tickets to a level that AI actually delivers a reasonable pull request, many corporate managers will seriously wonder why keep the offshoring team around.
It seems like the Holy Grail here has become: "A business is one person, the CEO, sitting at his desk doing deals and directing virtual and physical agents to do accounting, run factories, manage R&D, run marketing campaigns, everything." That's it. A single CEO, (maybe) a lawyer, and a big AI/robotics bill = every business. No pesky employees to pay. That's the ultimate end game here, that's what these guys want. Is that what we want?
So far, automation has only ever increased the need for software development. Jevons Paradox plus the recursive nature of software means that there's always more stuff to do.
The real threats to our profession are things like climate change, extreme wealth concentration, political instability, cultural regression and so on. It's the stuff that software stands on that one should worry about, not the stuff that it builds towards.
Maybe I’m not think big picture enough… but have you ever tried using generative AI (i.e., a transformer) to create a circuit schematic? They fail miserably. Worse than Chat GPT-2 at generating text.
The current SOTA models can do some impressive things, in certain domains. But running a business is way more than generating JavaScript.
The way I see it, only some jobs will be impacted by generative AI in the near term. Not replaced, augmented.
Aider is definitely in the same camp. Last time I checked, they weren't optimizing for the full "agent infinitely looping until completion" usecase, and didn't have MCP support.
But it's 100% the same class of tool and the awesome part of the unixy model is hopefully agents can be substituted in for each other in your pipeline for whichever one is better for the usecase, just like models are interoperable.
The vision of submitting a feature request and receiving a ready-to-review PR is equally compelling and horrifying from the standpoint of strategy management.
Like Anthropic and most big tech companies, they don't want to show off the best until they need to. They used to stockpile some cool features, and they have time to think about their strategy. But now I feel like they are in a rush to show off everything and I'm worried whether the management has time to think about the big picture.
Setting aside predictions about the future and what is best for humanity and all that for a moment this is just such a bummer on a personal level. My whole job would become the worst parts of my job.
(please pardon the self-promotion) This is exactly what my product https://cheepcode.com does (connects to your Linear/Jira/etc and submits PRs to GitHub) - I agree that’s the golden state, and that’s why I’m rushing to get out of private beta as fast as I can* :) It’s a bootstrapped operation right now which limits my speed a bit but this is the vision I’ve been working towards for the past few months.
*I have a few more safety/scalability changes to make but expecting public launch in a few weeks!
> The "golden" end state of coding agents is that you give it a Feature Request (EG Jira ticket), and it gives you a PR to review and give feedback on. Cursor, windsurf, etc, are dead ends in that sense as they are local editors, and can not be in CI.
Isn’t that effectively the promise of the most recently released OpenAI codex?
From the reviews I’ve been able to find so far though, quality of output is ehh.
played around with connecting https://github.com/eyaltoledano/claude-task-master via mcp to create a prd which basically replaces the ticket grooming process and then executing it with claude code creating a branch named like the ticket and pushing after having created the unit tests and constant linting.
Claude Code is my favorite way to use LLMs for coding.
However I feel what we really need is to have an open source version of it where you can pass any model and also you can compare different models answers.
(Aider and other alternatives really doesn't feel as good to use as Claude Code)
I know this is not what anthropic would want to do as it removes their moat, but as a consumer I just want the best model and not be tied to an ecosystem. (Which I imagine is the largest fear of LLM model providers)
OpenAI codex is probably the closest to what you're talking about, its open source and you can use models from any provider. It's not as good as claude code right now but I bet it wont take long for them to catch up.
Aider has had support for Python and shell scripting [0] for a long time.
I made a screencast [1] recently that included ad-hoc bash scripting aider as part of the effort to add support for 130 new programming languages. It may give a flavor for how powerful this approach can be.
Freaking love Aider. MCPs are supported soon as well. Testing a development branch. Then you can actually develop end to end using PR, tickets etc using models you trust.
Maybe I'm holding it wrong, but I can easily spend $20+ using Claude Code for 2 hours. I've stopped using it because it was too expensive for my personal projects.
I’ve really enjoyed the recent latent space podcasts. I don’t think there is any person†/podcast (or perhaps other content) approaching your general output while maintaining the high SNR. I am continually amazed at the volume and value of public work you’re producing over the last (half?) decade while still growing various businesses. I hope others can find similar productivity gradients. I know you roughly share what works for you but it is not so easy to reproduce.
thanks man, this was nice to read :) idk if it helps but my principles (tm) are here http://learninpublic.org/
i do feel like SNR * quantity could be higher, but its still a challenge to even keep it where it is today. my work life balance/stress levels aren't the best and everyone expects everything from me.
If I was making an AI code assistant, the last thing I would do is to lock it in to a particular foundation model provider.
The only possible way for this to be a successful offering is if we have just now reached a plateau of model effectiveness and all foundation models will now trend towards having almost identical performance and capabilities, with integrators choosing based on small niceties, like having a familiar SDK.
Other than the command/arguments there isn't much locking you in. It's just input/output. Swap it out for something else or simply wrap it. There's not much going on here.
Claude Code could already be used in non-interactive mode, and by extension it could be integrated into other apps in the same manner as any other UNIX command line utility.
This SDK currently supports only command line usage. Isn't that just what we already had?
I don't understand what's actually new here. What am I missing?
I would also recommend Codebuff (https://www.codebuff.com/), a great CLI code assistant comparable to Claude Code, which can save a lot on token costs.
(I am not affiliated with this project, just a user.)
> You may not access or use, or help another person to access or use, our Services in the following ways:
> 2. To develop any products or services that compete with our Services, including to develop or train any artificial intelligence or machine learning algorithms or models or resell the Services.
Can somebody please tell me what software product or service doesn’t compete with general intelligence?
Imagine selling intelligence with a legal term that, under strict interpretation, says you’re not allowed to use it for anything.
Is it so vague it’s unenforceable?
How do we own the output if we can’t use it to compete with a general intelligence?
Is it just a “lol nerd no one cares about the legal terms” thing? If no one cares then why would they have a blanket prohibition on using the service ?
We’re supposed to accept liability to lose a lawsuit just to accept their slop? So many questions
Readit News
The "golden" end state of coding agents is that you give it a Feature Request (EG Jira ticket), and it gives you a PR to review and give feedback on. Cursor, windsurf, etc, are dead ends in that sense as they are local editors, and can not be in CI.
If you are tooling your codebase for optimal AI usage (Rules, MCP, etc), you should target a technology that can bridge the gap to headless usage. The fact Claude Code can trivially be used as part of automation through the tools means it's now the default way I thinking about coding agents (Codex, the npm package, is the same).
Disclaimer, I focus on helping companies tool their codebases for optimal agent usage, so I might have a bias here to easily configurable tools.
That's what I want and look forward one day
I hate using voice for anything. I hate getting voice messages, I hate creating them. I get cold sweats just thinking about having to direct 10 AI Agents via voice. Just give me a keyboard and a bunch of screens, thanks.
I'm not saying "ban propietary LLMs", I'm saying: hackers (the ones that used to read sites like this) should have as their main tools free and open source ones.
Yes, because hardware and electricity aren't free.
I literally DO pay for every command. I just don't get an itemized bill so there's no transparency about it. Instead, I made some lump-sum hardware payment which is amortized over the total usage I get out of it, plus some marginal increase in my monthly electric bill when I use it.
I was doing this with Cursor and MCPs. Got about a full day of this before I was rate limited and dropped to the slowest, dumbest model. I’ve done it with Claude too and quickly exhaust my rate limits. And the PRs are only “good to go” about 25% of the time, and it’s often faster to just do it right than find out where the AI screwed up.
I see your point but in the other hand how depressing to be left only with the most soul crushing part of software entering - the Jira ticket.
I understand the craft of code itself is what some people love though!
The real threats to our profession are things like climate change, extreme wealth concentration, political instability, cultural regression and so on. It's the stuff that software stands on that one should worry about, not the stuff that it builds towards.
The current SOTA models can do some impressive things, in certain domains. But running a business is way more than generating JavaScript.
The way I see it, only some jobs will be impacted by generative AI in the near term. Not replaced, augmented.
No, any team.
Put the Aider CLI into a GitHub action that's triggered by an issue creation and you're good to go.
But it's 100% the same class of tool and the awesome part of the unixy model is hopefully agents can be substituted in for each other in your pipeline for whichever one is better for the usecase, just like models are interoperable.
Like Anthropic and most big tech companies, they don't want to show off the best until they need to. They used to stockpile some cool features, and they have time to think about their strategy. But now I feel like they are in a rush to show off everything and I'm worried whether the management has time to think about the big picture.
Deleted Comment
*I have a few more safety/scalability changes to make but expecting public launch in a few weeks!
Isn’t that effectively the promise of the most recently released OpenAI codex?
From the reviews I’ve been able to find so far though, quality of output is ehh.
I bias a bit to wanting the agent to be a pluggable component into a flow I own, rather than a platform in a box.
It'll be interesting to see where the different value props/use cases of a Delvin/v0 vs a Codex Cloud vs Claude Code/Codex CLI vs Cursor land.
However I feel what we really need is to have an open source version of it where you can pass any model and also you can compare different models answers.
(Aider and other alternatives really doesn't feel as good to use as Claude Code)
I know this is not what anthropic would want to do as it removes their moat, but as a consumer I just want the best model and not be tied to an ecosystem. (Which I imagine is the largest fear of LLM model providers)
https://github.com/openai/codex/tree/main
It's still under development but looks promising.
What does Claude Code do better than Aider?
[0] https://aider.chat/docs/scripting.html
[1] https://aider.chat/docs/recordings/tree-sitter-language-pack...
I don't really want it committing and stuff, i mostly like the UX of Claude Code. Thoughts?
Add a file to your repo and you can talk to any model via issues.
you can skim the transcript but some personal highlights:
- anthropic employees, with unlimited claude, average to $6/day of usage
- headless claude code as a "linux" utility that you use everywhere in CI is pretty compelling
- claude code as a user extensible platform
- future roadmap of claude code: sandboxing, branching, planning
- sonnet 3.7 as a persistent, agentic model
From the link:
"Apparently, there are some engineers inside of Anthropic that have spent >$1,000 in one day!"
The question is what is the P50, P75, and P95 spend per employee?
https://news.ycombinator.com/item?id=43737060
† simonw, gwern
i do feel like SNR * quantity could be higher, but its still a challenge to even keep it where it is today. my work life balance/stress levels aren't the best and everyone expects everything from me.
Deleted Comment
The only possible way for this to be a successful offering is if we have just now reached a plateau of model effectiveness and all foundation models will now trend towards having almost identical performance and capabilities, with integrators choosing based on small niceties, like having a familiar SDK.
At this point Claude Code is a software differentiator in the agent coding space.
I am building things related to AI code assistants - we were hacking ways to integrate Claude Code - it was the first thing we wanted to build around.
It's too early to care about lock in.
Need the best, will only build around the best.
This SDK currently supports only command line usage. Isn't that just what we already had?
I don't understand what's actually new here. What am I missing?
(I am not affiliated with this project, just a user.)
Can somebody please tell me what software product or service doesn’t compete with general intelligence?
Imagine selling intelligence with a legal term that, under strict interpretation, says you’re not allowed to use it for anything.
Is it so vague it’s unenforceable?
How do we own the output if we can’t use it to compete with a general intelligence?
Is it just a “lol nerd no one cares about the legal terms” thing? If no one cares then why would they have a blanket prohibition on using the service ?
We’re supposed to accept liability to lose a lawsuit just to accept their slop? So many questions