This is of course subjective, but I would give a lot to have an alternative to Claude Code and the Claude models, but there just isn't anything comparable that works well in an integrated manner for agentic coding.
It's not like I haven't tried. Gemini CLI is still trash (it's probably a bit better now, but I still can't see the edits it proposes, well, etc.). I tried OpenCode, the whole experience was frustrating: the models give up mid-task, they run rampant with actions, the CLI does not offer the level of control and customization Claude Code offers, etc.
I've also tried the other major tools: Codex, Cursor, Cline, Aider, and others, nothing works for me. You are surprised people stick to Claude Code, I am surprised people bother with the other tools.
Maybe it has something to do with how I use the agentic tools: I use the CLI almost exclusively, rarely using the IDE (unless I want to actually code myself). I also almost always approve each and every edit. As such, my number one concern is for the tool to provide me with proper control in a simple and reliable manner: I want a rich permission system that works, and I want to see each proposed edit very clearly in an ergonomic diff format. I want to be able to type, recall, and edit my commands easily too. These are things Claude Code excels at that the other just don't.
The best I've been able to do is to use third-party routers to enable me use Claude Code with almost-SOTA models, and this is the approach that shows the most promise. I'd hate to be beholden to Anthropic's shenanigans.
I think what will eventually help is something I call AI-discipline. LLMs are a tool, not more, no less. Just like we now recognize unbridled use of mobile phones to be a mental health issue, causing some to strictly limit their use, I think we will eventually recognize that the best use of LLMs is found by being judicious and intentional.
When I first started dabbling in the use of LLMs for coding, I almost went overboard trying to build all kinds of tools to maximize their use: parallel autonomous worktree-based agents, secure sandboxing for agents to do as they like, etc.
I now find it much more effective to use LLMs in a target and minimalist manner. I still architecturally important and tricky code by hand, using LLMs to do several review passes. When I do write code with LLMs, I almost never allow them to do it without me in the loop, approving every single edit. I limit the number of simultaneous sessions I manage to at most 3 or 4. Sometimes, I take a break of a few days from using LLMs (and ofter from writing any code at all), and just think and update the specs of the project(s) I'm working on at a high level, to ensure I not doing busy-work in the wrong direction.
I don't think I'm missing anything by this approach. If anything, I think I am more productive.
As someone mentioned on this thread, I can also easily out-engineer Claude Opus, lol its not even close.
Note that I'm not talking about the low-level grunt work (and even with that, its just that it is tedious and time-consuming, but if I had enough time to read through all the docs and stuff, I will almost always produce grunt code of much higher quality).
But I'm more talking about architecture, the stuff of proper higher level engineering. I use Claude Opus all the time, and I cannot even count how many times I've had to redirect its approach that was obviously betraying a complete lack of seeing the big picture, or some egregiously smelly architectural approach.
Also, expressive typing. I use mostly TypeScript, and it will often give up when I try to push it beyond a certain point, and resort to using "any". Then I have to step and do the job myself.
Not even close. There are still tons of architectural design issues that I'd find it completely useless at, tons of subtle issues it won't notice.
I never run agents by themselves; every single edit they do is approved by me. And, I've lost track of the innumerable times I've had to step in and redirect them (including Opus) to an objectively better approach. I probably should keep a log of all that, for the sake of posterity.
I'll grant you that for basic implementation of a detailed and well-specced design, it is capable.
> Iran has been helping arm Russia, and evade sanctions
Not sure this bit makes sense. The current occupant of the white house is firmly in the camp of Russia; that is plain for all to see. In fact, the current war is driving up oil prices and thus actually helps Putin, which may be intended. They are also lifting sanction on Russia.
He's not exactly in their camp. I hate to invoke Hanlon's Razor because a lot of what he does is malice rather than stupidity, but on Russia he has been inconsistent in a way that doesn't suggest that he's just following orders.
If he really wanted to just give up Ukraine, he could have. It's not as if he's afraid to piss off our allies. There's no 4D chess going on.
He likes Russia. He admires their strongman. He wants to be one too. He doesn't want to just give up Ukraine, but neither does he care what happens to it. Like Putin he's happy just to make everyone else deal with his random behavior.
He has a ton of advisers with conflicting goals. Some of them really do want to take down Russia, or prolong its involvement in Ukraine to distract them. Any explanation for this war is the sum of those conflicting advisors.
It's often said that he does whatever the last person in the room wants. I don't think it's quite true, but it's not way to understand how his contradictory actions come about.
Careful there. I've resolved (and succeeded somewhat) to tone down my swearing at the LLMs, because, even though the are not sentient, developing such a habit, I suspect, has a way to bleeding into your actual speech in the real world
To be honest “no dummy” is how you would swear at a 4-year-old.
I often use things like: “I’ve told you no a bilion times, you useless piece of shit”, or “what goes through your stipid ass brain, you headless moron”
I am in full Westworld mode.
But at least when that thing gets me fired for being way faster at coding than I am, at least I’d haves that much frustration less. Maybe?
It does. But then, it's how i talk to myself. More generally, it's how i talk to people i trust the most. I swear curse and insult, it seems to shock people if they see me do it (to the llm). If i ask claude or chatgpt to summarize the tone and demeanor of my interactions, however, it replies "playful" which is how im actually using the "insults".
Politeness requires a level
of cultural intuition to translate into effective action at best, and is passive aggressive at worst. I insult my llm, and myself, constantly while coding. It's direct, and fun. When the llm insults me back it is even more fun.
With my colleagues i (try to) go back to being polite and die a little inside. its more fun to be myself. maybe its also why i enjoy ai coding more than some of my peers seem to.
For example, sometimes it outputs in markdown, without being asked to (e.g. "**13**" instead of "13"), even when asked to respond with a number only.
This might be fine in a chat-environment, but not in a workflow, agentic use-case or tool usage.
Yes, it can be enforced via structured output, but in a string field from a structured output you might still want to enforce a specific natural-language response format, which can't be defined by a schema.
No, you can't do real work on a $350 windows machine. No way such a setup is suitable for anything beyond browsing a tab or two and connecting to servers using SSH.
And, the whole shittiness of the experience will even distract you attempting real work: the horrible touchpad, the bad screen, the forced windows updates when you trying to start the machine to do something urgent, ads in Windows, the lack of proper programmability of Windows (unless you use WSL).... Add the fact that the toy is likely to break in a year or two. These issue exist on far more expensive Windows machines, how much more a $350 machine.
Leaving Windows machines and OS behind for more than a decade has been a continuing breath of fresh air. I have several issues with the Apple devices and macOS (as I have with Linux too), but on the whole they are far better than Windows. The only good thing about Windows that I miss on Macs is the file explorer and window management, not sure why Apple stubbornly refuses to copy those.
A lot of $350-ish Windows machines also don’t have SSDs but instead eMMC storage, which is dog slow and will make modern SSD-mandatory Windows feel even more awful to use.
If Windows/Linux/x86 is non-negotiable and that’s your budget, I would never in a million years recommend anything brand new. This is when you go pick up a $350 used midrange ThinkPad on eBay. It won’t outperform a Neo in terms of CPU and battery life but I guarantee it’ll be a better experience than the garbage routinely sold at this price point.
Of course you can. You can do real work on an $80 Amazon Fire. Yes, some things will be potentially impossible or frustrating but that's also true of the MacBook Neo, just a bit higher of a bar. A lot of this also depends on your definition of "real work".
$350 USD can get you a decent laptop with a SSD, 16GB RAM and something like an Intel N100 or N95. And they pretty comparable to a decent Intel Skylake CPU which are still pretty usable.
Yes, the Neo has a faster CPU but it also has less RAM and less storage and costs more and has less ports. Besides ray traced games what can the Neo do that the others can't? They'll take longer but they'll get there.
And if you're willing to go used? That $350 goes a lot further.
> Yes, the Neo has a faster CPU but it also has less RAM and less storage and costs more and has less ports.
8GB on Apple Silicon is far better than 16 GB on Wintel, and I don't event trust the quality of 16GB of RAM on a bottom of the barrel Windows machine.
Would you prefer a machine that is still good 7 years from now with less ports, or one with more ports that you have to replace in 2 years? Yes it is more expensive now, but over 7 years it is an absolute bargain.
16 GB physical RAM is just better. Apple isn't magic. Gimme a break. Both devices have SSDs for fast swapping and have RAM compression. You can't spin up a VM that has 8GB RAM on the Neo, you can't load a large spreadsheet or do a decently sized digital painting. I could maybe buy a claim that 8GB is better on Mac than 8GB on Windows.
Why would you have to replace it in 2 years? How do we know Apple will even be offering updates to Neo in 7 years? Will 8GB still be usable in 7 years really? 8GB is barely on the fence already.
I wouldn't be surprised if Apple drops the Neo from software support in less than 7 years.
The ThinkBook 14 Gen 6 at Costco for $380 has a single thread passmark score of 2800. The laptop I use to develop most of my SaaS products, with IDEs and claude open etc, has a score of 2000. I run Linux, but win10 iot runs fine on it too.
> No, you can't do real work on a $350 windows machine.
Sigh. I mean, even absent the obvious answers[1], that's just wrong anyway. You're being a snob. Want to run WSL? Run WSL. Want to run vscode natively? Ditto. Put it on a cheap TV and run your graphical layout and 3D modelling work. I mean, obviously it does all that stuff. OBVIOUSLY, because that stuff is all cheap and easy.
All the complaining you're doing is about preference, not capability. You're being a snob. Which is hardly weird, we're all snobs about something.
But snobs aren't going to buy the Neo either. Again, the business question here is whether the $350 junk users can be convinced to be snobs for $600.
[1] "Put Linux on it", "All of your stuff is in the cloud anyway", "It's still a thousand times faster than the machine on which I did my best work", etc...
You mean that machine from 30 years ago that was running 30 year old software that has nothing in common with today’s development? And how well does Linux run on 4GB?
That's a 16G windows box which will happily run multiple VMs for whatever your deployment environment is, something the Neo is actually going to struggle with. The Jasper Lake CPU is indeed awfully slow, but again for routine "dev" tasks that's just not a limit.
You would obviously refuse out of taste, but if you were actually forced to use this machine to do your job... you absolutely could.
It's not like I haven't tried. Gemini CLI is still trash (it's probably a bit better now, but I still can't see the edits it proposes, well, etc.). I tried OpenCode, the whole experience was frustrating: the models give up mid-task, they run rampant with actions, the CLI does not offer the level of control and customization Claude Code offers, etc.
I've also tried the other major tools: Codex, Cursor, Cline, Aider, and others, nothing works for me. You are surprised people stick to Claude Code, I am surprised people bother with the other tools.
Maybe it has something to do with how I use the agentic tools: I use the CLI almost exclusively, rarely using the IDE (unless I want to actually code myself). I also almost always approve each and every edit. As such, my number one concern is for the tool to provide me with proper control in a simple and reliable manner: I want a rich permission system that works, and I want to see each proposed edit very clearly in an ergonomic diff format. I want to be able to type, recall, and edit my commands easily too. These are things Claude Code excels at that the other just don't.
The best I've been able to do is to use third-party routers to enable me use Claude Code with almost-SOTA models, and this is the approach that shows the most promise. I'd hate to be beholden to Anthropic's shenanigans.
reply