Workflow in Cursor is actually awesome - I'm a little outdated in how I use it - I still establish goals/objectives, rather than managing the loop which does so - but if you can think broadly enough - I find it's pretty efficient.
Key things I like about Cursor (and I recognize I'm dating myself a bit here): - Plan Mode is really solid - I shift-tab, have it go create the plan using whatever insanely expensive SOTA model is available - I will usually spend 5-10 minutes on the Plan - review it, maybe even tweak it a little. (though 90% of the time it's fine out of the gate)
- Ability to select any model for every task - I'll switch between Opus 4.8 High/xHigh/... I'll even switch to 1M context for the planning phase upfront.
- It does an *excellent* job managing permissions and looping the agents and spinning up sub-agents for you - you set the goal, run the plan mode - and then let it churn for however long is required - pretty common to have a 30-45 minute run and come back to a fully created/tested product.
The nice thing about Cursor (and honestly Claude Code, Codex) - there isn't really any "prompt engineering" involved. You just say, "Go Build me x - it should have y,z features - and build it in golang for me" - and that's it - the 3-4 page Plan comes back - usually pretty credible - and then you click "build.".You should make an experiment; take someone who never used any LLMs or agents, and tell them to use it for the first time in front of you, and tell them to build something like a calculator program or whatnot. Bonus points if they're ICs or at least not-managers.
I think there is a lot us engineers take for granted, when it comes to communicating via text, how to state things clearly and what we think/reason when we read things. A lot of people don't have those "skills" innate, and the first time they use LLMs, they basically don't know how to interact with them, until they realize what they're able to do and not. Then they also learn what to say to steer the model into the right way, this is quite literally a "prompt engineering" skill they're now learning.
It's a big part of what, in my experience, is separating the very good engineer from the iffy one: Do you have a good mental model, and can you put yourself in the shoes of people sitting in a different mental model? It makes you a better dev, and even more so when it comes to AI tools, which have their own kind of alien brain.
I like to chat with Claude about how to approach a given problem, bring in extra context, etc, before even really drafting up a plan, while other people dive into implementation immediately and go on wild goose chases.
90% of the time we end up in the same place in roughly the same amount of time, and there are obviously tradeoffs to spending more time planning vs implementing. I'm oversimplifying as well.
Until now, the actual act of writing code: terminology, syntax, etc. was a significant hurdle, and that underlying mindset was a very useful, but missing in a surprisingly large number of developers, skill.
Now with LLMs doing the work of "translate this into code," increasingly the only thing that matters is that exact ability. And developers that don't have it or can't develop it won't be developers for long.
But - I don't know if it was April, or May - but very recently - the coding harnesses paired with decent SOTA models like Opus 4.8/GPT 5.5 - just started showing a lot more consistency, and completeness, and sometimes downright clever behavior - that they started to become way more useful.
Just one out of hundred+ examples - I gave Claude Code (Opus 4.8 High) a complex task that involved consul, vault - but I had neglected to give it sandbox permission to download from hashicorp.com. So - it created a entire test harness that simulated both the behavior of Vault and Consul - created all it's test cases, verified that they passed - and when I came back 40 minutes later said that it was all done.
It's test harnesses so accurately simulated the behavior of Vault/Consul - that on first try - no refactoring whatsoever - all of the protobuf/AESGCM/API behavior (that has varied significantly between versions) - worked.
This was something that would have taken me, someone super super familiar with the code and tools and APIs - a minimum of 3 solid days of work - and that would likely involve hundreds of attempts and refactors as I unwound all the weird encryption and packaging layers. It zero-shotted a full solution without having an API to test against
If these agents actually have an actual test-harness - It's honestly hard to imagine what they can't do - subject only to imagination and budget at this point.
Speaking personally - something changed Between January and, Let's say May - in which instead of seeing these things as mostly interesting technology demonstration, in which the flaws outweighed the benefits - I now genuinely think they are the future of programming. I'm dubious that I'll write much software manually in the future - beyond what I do for personal pleasure.
Some people "got" LLMs back in 2022, others needed it to evolve a bit.
It's not unlike computers. I started using them back in the 90s and absolutely nobody I knew was interested, while today everyone carries one in their pockets...
1) The data they have flowing through the system that enabled them to build composer (which is much better than stock kimi 2.5) and is presumably allowing the training of a new model on space Xs compute.
2) Cursors new 'github' replacement.
3) Enterprise sales/traction
If you look at all of these together, it's not implausible that they end up mostly 'owning' coding in 5 years time. If they replace GitHub with something more compatible with agentic coding and bring it into their whole ecosystem providing cloud and local agents, PR review and own frontier coding model.
It's specialised vs 'borg' isn't it. One way of thinking is that the world is owned by Anthropic/OpenAI and coding is just one of many things their model and software does. Another view is we have a 'coding with LLMs' company that specialises in this field of endeavour. Hard to say which wins, but I think they have a shot.
Personally my only objection to cursor is that it's more expensive. That's it, otherwise it is great to be able to choose say GPT-5.5 when I want to work on backend and Opus when I want to work on front end. Great to have PR review built in. If they were able to get composer 3 to as good as GPT5.5 / fable at the price of composer 2.5 they'd be winning on price again.
They really need to change their trajectory then?
And regardless being owned by xAI, a failed AI company which turned into a datacentre operator probably won't help them to achieve that.
> Hard to say which wins, but I think they have a shot.
The market for "coding harnesses" and "AI IDEs" is already oversaturated and they are effectively a commodity at this point, you can use any of them with any provider more or less interchangeably.
I think near unlimited access to compute is exactly what they need to train a frontier level coding model and serve it cheaply and profitably.
> The market for "coding harnesses" and "AI IDEs" is already oversaturated
I think my entire point was that it's not just a AI IDE. It's a coding focused model (currently Composer 2.5, soon hopefully something better), a Github Replacement, PR review/Bug Bot, Cloud Agents and so on and so forth. It's a ecosystem. An enterprise signs a MSA with you and gets everything they need all in one place.
Yes because Grok failed and they now have "unlimited" compute they can sell to other. I mean you are right that if they did X, Y and Z they could be very successful but their is no indication that might happen. In any meaningfully way seems like Cursor has peaked a while ago.
> An enterprise
Well either they are the type of companies which just buys whatever Microsoft is selling OR they let their developers to mostly pick what they feel is the best tool for the job on their won. I don't think there is that much in between (and its a cutthroat market e.g. GitLab)
> a Github Replacement, PR review/Bug Bot, Cloud Agents
Those things are a dime a dozen, you can vibe code them in weeks/months and there plenty of options on the market already. Well not Github of course, but there are various reason for that which have little to do with product quality and features (not that I think there are many companies which could build a meaningful GH replacement in a realistic time period despite its many flaws).
I just don't really see a huge income stream for dev tools companies (just like there never was) they can skim of something from the top by reselling AI models (generally at zero or negative margins..) but that's not the most lucrative business model when you have no real moot.
So what? How much money is that making?
the failure is in converting regular people into actual ai product consumers. Companies are realising that the money is not in regular consumers but in enterprise and they are not considering grok as a serious alternative.
if anything, the name, the branding and the x/twitter affiliation has hurt adoption from money makers rather than help it.
so yes, people know it, but no one is willing to pay for it
Does it mean they are out of the race? I have no idea, but things don't look great.
Yes and no. I've used a few different harnesses with closed and open models and there is definitely something going on that makes some harnesses work better than others. Many of the differences are hard to pin down and some are things people don't care about. But I wouldn't say they are commodified just yet.
1. Memory use. I have colleagues complaining that Clause Code uses several GB of memory. Meanwhile I haven't heard about that regarding codex or goose, or even opencode for that matter.
2. Suitability for local models. When you use Anthropic models, you use Anthropic as a provider. They can have software between the model and your harness that will fix issues with the model. One notable thing that even the best open weights models struggle with is broken tool calls. There is a lot that a harness can do to fix broken tool calls when working with a straight up ollama running a raw GGUF file.
3. Ease of use with non mainstream models. OpenCode has GREAT coverage of models/providers. Goose, less so as it relies on people to set up their own anthropic or openai compatability settings. e.g. Zed doesn't let you use Z.ai (which, if you speak British English, sounds ironic because "zed ai" isn't directly supported by Zed the editor).
4. Worktree support. Opencode and probably all the TUI harnesses works in a local directory - so you need the terminal to be in the worktree. Zed, however, works centrally on your git repo and tracks the worktrees so you can bounce around your work in a single window.
Of these, '2' is maybe the most important one but also the hardest to pin down as a feature. '3' is a one time cost. Of course '1' could be a blocker for someone using a macbook air or neo.
I'm not a gitlab user, just learning it, so I can't say how half baked they are or not.
At a high level though it seems like a huge step forward than GitHub
This is the same data they used to finetune Kimi K2.5 to make their newer Composer models, which benchmark substantially better than Kimi K2.5.
I've heard they also want to build their own base models, which will also benefit from their large amount of high-quality training data. Which will solve Grok's model quality problem.
This is all unsourced conjecture of course. But it's what I've heard.
So it's not really fair to call it "fine tuning", it's an important part of building a coding model in 2026, and cursor have done a pretty good job with Composer
Maybe because SpaceX paid with monopoly money (all stock deal)?
xAI overbuilt their data centers - they can't find paying customers for them, that's the reason they made deals with other companies like Google to use their own datacenters.
Cursor has the opposite problem of not having enough capacity. So this works well for them together.
Weather it's worth it - if you beleive that AI will solve every problem then having a piece of the pie early on might be worth it.
Remember how when google bought youtube for 1.65 billions people thought they are crazy? Or when facebook bought instagram.
60B is a crazy number but might be worth it for someone fighting for world dominance :)
xai is on the line to delivery capacity they already sold to Google and most analysts think they are 50/50 on actually meeting it.
the only proof they have capacity is that musk claims all the money they are burning is going to datacenters and gpu (mostly because if he put it on anything else the lie would be obvious)
Or are they paying for talent? It seems like xAI is sorely lacking in talent, most likely due to the CEO and folks' aversion to him. By throwing around some SpaceX monopoly money he can trap some talent with retention clauses and try to invigorate his failed AI business.
I think these are good examples: in both of those cases the buyer had a plan to monetize.
If you are a user of Cursor, expect to pay more for it or switch.
For anyone that doesn't get the reference, please start here [1].
1: https://www.chinatalk.media/p/how-to-buy-cheap-claude-tokens...
From a valuation model - $4B ARR with rapid growth, and the ability to shift traffic to internal models (honestly, massive amount of the time "composer" - their internal model is fine, and obviously going to get better). Say 17x Multiple which isn't unheard for a rapidly growing Startup with solid future structural profit elements (moving to internal model) - that gets you to $68B.
Being able to compare outcomes for workflows involving competitors will obviously be v v v v useful.
If you resell something worth $5 for $5 while having to pay for R&D and operating expenses that's not exactly comparable with a company that's selling actual products.
> Say 17x Multiple
On an extremely low margin business it is, yet again that wouldn't be the stupidest thing in today's market.
But there are many model agnostic harnesses out there: OpenCode, Roo, Cline, and many others. And even Claude Code can be setup to use non-Anthropic models.
Open routers prices are no different than cursors and you can use any harness you want.
Big brain, small brains? Hmmm
Too much friction though, with Cursor its out of the box.
We’re in the new era where startups boast about and bought based on revenue and not on just a number of users with unclear path to monetizing as it had been for the previous couple decades.
We can also note that we see Thrive Capital (Kushner) again in a win.
It can't as long as there is plenty of AI without it.
The real differentiatior is that if $60B today turns out to be all thrown away in a worst-case scenario, it would be easily more affordable and there would be less negative impact than $47B at the time if it was all thrown away on Twitter.
I don't see how these systems can have the ability to be effectively expressive about all of the minutia, and not have all of the various different possible expressions lead to vastly different outcomes.
For example - I might (real world example from this morning):
"Create a script that installs hashicorp vault and consul, store the data on consul. Then create ahelper script that will fill the vault server with sample data. Add HTTPS support. Now write a framework that reads and decrypts the encrypted data in consul. Support old (pre 1.3) and new (post 1.3 vault). "
That generates a 6 page plan using Opus 4.8 w/1mm context, including notes on what to prioritize, what format to create the scripts in, etc... (My cursor guidance already has a couple months of hints as to what I want in terms of scaffolding unit tests, canonical linux, performance, security, etc...)
That 6 page plan is the "Prompt" - but it's entirely generated by Cursor/Opus. It's there to tweak if you want to emphasize, or provide some taste - but, honestly - it probably does a better job than I would - so ~90% of the time I just accept the plan as is.
What actually matters is the ability to communicate well in general, not anything LLM-specific. Being able to state what you want clearly and unambiguously, and having a sense for what additional information you need to dump, even when the other side claims they already have everything they need.
Cursor does have it's own model - it's a heavily reworked version of KimiK2, called "composer" - that I use a lot of the time when I have fairly straightforward tasks that don't require a lot of exploration or independent thought. Lot cheaper - the Input/CacheWrite/CacheRead/Output costs of Opus 4.8 are $5/$6.25/$0.5/$25 per mm tokens, vs $0.5/-/$0.2/$2.5.
So your workflow now looks like mine except I prefer a different editor and only use the latest and greatest model so Cursor basically offers nothing over Codex.
I disagree about prompt engineering, but it's one of those things that probably varies because of what language you use, what problems you solve, and the degree to which you care about the output. Unless I'm writing tests, I keep AI on a very short leash because I'm writing critical code used by a very large number of users. I have noticed big differences in output quality depending on how I steer AI. Without steering, it will happily leave in dead code, change the use of variables so they need to be renamed, assume or fail to assume invariants, etc. As I said in another comment, I think we won't need to do that for very much longer, but right now it seems essential.
Over time - I expect Composer will be cheaper than Opus 4.8 - but the nice thing about Cursor - you can flick between models.
And (this is purely a personal thing) - I really like the extensive collection of "Plans" that cursor tracks - there isn't really a similar thing in Claude Code - but I really like the Claude.AI interface for everything else. It's also a much better general knowledge agent - the Cursor Chat interface isn't as nice.
Maybe you haven’t looked deeper into what modern Claude can do?
It's not perfect, btw - to some degree you are at the mercy of which models they support - currently only 27 from Gemini, OpenAI, Anthropic, Grok, and Kimi (Just K2.5) - presumably because they have commercial arrangements with them. The "Bring your own Model" model requires you plunk in your API key - which sucks. And only one at a time.
To the best of my Knowledge, Claude Code only supports one model at a time if it's not one from Anthropic (which will use the the entire suite of Anthropic Models depending on the task) - and you have to override it to a single model with an environment variable at startup - no ability to flick between models from task to task.
Depending on your workflow - you can save 70-90% on costs just by chosing a reasonable model for really extensive tasks that don't require thinking, max context, etc....
The end point may squeeze quality of operations at the subminute time span for ground control environment seriously launching Starship rockets one an hour, for example.
Some of the issues (off the top of my head):
- Note - that my "product" was about 3,000 lines of code - so tiny. But https://metr.org/ should give you some insight into the complexity the models are capable of.
- you have to be able to imagine the product. If I have the time, and energy, to imagine what I want - the model will build it. Here is an example of a much better programmer than I and something he wanted built - https://www.boatbomber.com/blog/claude-fable-5
- These are the first drafts. On average - any complex system needs about 10 years and at least 1000 active and enthusiastic about reporting users to really get robust code. Writing if via LLM doesn't (at least so far in my experience) help that much in reducing bugs if you were previously following any semblance of TDD. Lots of bugs in the code - the products you listed above have literally tens of millions of years of user experiences and bug reports that got them to where they are today. No silver bullet yet - just faster, less effort - and it enables non-technical people to create (still buggy) products.
What a world we live in - "dating oneself" is measured in weeks/months! :)
What you're describing seems like a workflow for building toys only. There's currently no reality in which someone would actually know what the y,z features are before making them. A plan generated in 5min would likely suggest a suboptimal solution compared to what a good solution would look like (which might take a year or two to figure out, for a human, so still a week or so for SOTA models if at all possible). Building something in golang is cute, but hard to be convinced until more novel applications are being generated from prompts.
The data submitted by Cursor's users tho, that seems to be very valuable.
When I first used Cursor, I hadn't used any of the "Vibe Code" tools out there, so it was pretty neat to have an assistant directly tied to the editor.
Once I learned how to use Codex, I just used a tmux split with NeoVim and have the effect I wanted. I haven't felt compelled to use Cursor at work since.
What do you mean by that? What is happening in just over a year or so?
Closing this gap is done in the harness, either through Skills, user behaviour/prompts , Agents.md etc etc.
I think that this is an area worth investing time in, but it is indeed hard to know what the scope of this is.
My objective KPI: for the few days I was using Fable (18hr a day), it would frequently push back against my design ideas and propose alternatives -- and they almost always felt better to me. Back to Opus now, still 18hr days - and I dont think it disagreed with me meaningfully even once since Saturdy. I consider myself and old hand -- and i think Fable really didn't need me to be very specific in my prompts, it would have done a good job regardless, or even despite my prompting.
Of course whether this is the future is anyone's guess. Maybe we will experience a butlerian jihad and there won't be any prompting whatsoever for completely different reasons :-)
I am pissed off by people calling Cursor an IDE … Cursor is text editor with AI agents bolted on. I still like what agents do and how the context is managed in Cursor but it is far far away from proper IDE.
I started with Cursor back in the day, but switched to Claude Code and then Codex when Cursor got too expensive.
If price wasn't an issue, maybe I'd prefer Cursor only because I can easily switch between models. But that's it. I always disliked the "accept/reject" workflow in cursor, but that's probably optional nowadays I guess?
For example, sometimes Claude just obsessively reads files and goes on massive tangents. Then when I stop it and ask, "why are you doing that?", it kindly apologizes and admits it shouldn't have gone on a tangent.
The token burn if I don't stop it would be quite high.
Granted, this might be because I'm not giving it optimal prompt/negative-prompt instructions though.
If I reject, then the AI will struggle to modify just the parts I disagree with, if I accept, the AI will tend towards adding code rather than updating the bad code.
At that point copy paste without agentic coding tends to work much better.
Anthropic is going to offer better pricing using their agentic harness. Why pay more for less?
An IDE at this point is best as a tool for code review. They need to start building better code review tools.
Also https://xcancel.com/mitchellh/status/2066657032938442833#m
I really don't see IDE's going out of fashion anytime soon.
I tried having it write some tests today. It got very close to what I want, but picked a stupid set of input values (two fields that look independent that should only be used with related values). I thought about "how do I explain this" and then just went in and fixed it myself.
How is it easier to write "Okay, go back to testBlah and change xxx to yyy" versus clicking on XXX in the IDE and typing YYY by hand? Maybe if you had 500 faulty tests and were forbidden from using search-and-replace for some reason.
It makes sense when code generation is the limiting factor, but I end up with a lot of changes where the actual code delta is smaller than the necessary prompt to convince the bot to produce it.
Review each step as much as you care. These things take time so you can just do other stuff while it’s cooking.
With proper isolation of projects you can easily have multiple sessions in parallel. I frequently have 4 to 8 parallel Claude Code sessions, each with whole trees of agents reproducing, speccing, planning, implementing and reviewing things.
For common mistakes, you can make it remember things or rely on reviews.
What?
> An IDE at this point is best as a tool for code review.
I heard from a friend that most devs building serious stuff still write code. It's shocking but true. (No code review needed.)
claude code was seriously annoying with the flickering, maybe it's fixed now, I don't know.
cursor also has a (bad) cli if you need it, it seems it's mostly used to setup remote agents, but it does the job in a pinch.
Moved on to Zed (native Rust rendering) 2 weeks ago -> nothing flickers.
Sadly, with Fable 5 cutoff, I am actively exploring CC alternatives. Pi/OMP.sh works great as an agent (definitely better than CC). GPT is seemingly not as good as Opus, but with better agent and better skills, it probaly won't matter anyway. GPT lets you use any agent on Pro subscription.
Ctrl+c -> new tab -> `claude —resume` is deeply ingrained at this point.
Just what I said. They offer paid plans through their tool. Said paid plans are kind of a dark pattern where it’s not immediately obvious the models are training on your data. The harness is fine but that kind of business turns me off and I am usually pretty neutral about those sorts of things.
Most of that switching is automated (oh-my-openagent - defaults sub-tasks to different roles, so for example I use MiniMax for explorer tasks and GPT 5.5 for deep design & review tasks, and GLM 5.2 for general orchestrator & most coding). If I hit usage limits it switches to a backup for that task. I'm not sure Cursor authenticates with all the subscription coding plans from all those companies - but if it does it can't be doing it any better.
I run it in a sandbox and its not phoning home.
(FWIW Im mostly using python for OCR, LLM calls, data analysis..)
And I don't know what it is but it feels the less familiar you are with a terminal, the less skilled you tend to be.
Definitely not a 100% case. But has been common in my experience
You know what else is a good sign of them willing to dig in and understand stuff at a lower level ? If they dig in and understan stuff at a lower level. Let's judge people on what matters - the actual work and value they bring. Not status symbols like their IDE of choice or how fast they type into the keyboard.
Am I an outlier or do you just judge people for weird reasons? I’ve never seen an IDE person judge a terminal person, it’s always the other way around - what’s up with that?
Its just an anecdotal experience.
I think I'd generalize my post more to say the more often somebody reaches for the terminal, in my anecdotal experience the more proficient they tend to be.
VSCode it is based on is text editor.
AI features are great but it is not IDE.
ILE? Integrated Language Environment. The former caters to a specific language ecosystem. Text editors are polyglots by contrast.
If you call text editor that has terminal bolted on that’s not IDE. It is exactly the same as if you would open terminal on the side. Grepping/search with regex is also far from code navigation that is available in an IDE.
Debugging tools and code navigation in a proper IDE like Visual Studio or JetBrains Rider is totally different development experience than editor with terminal.
I also think your characterization about hobby/production fall short. My fastAPI and rails codebases anre definitely not a hobby.
Thanks for clarifying though. Interesting to hear your perspective.
That's not what I said?
What I said the was the line between hobby/production is use of the tools themselves. Doesn't matter if you use them via a terminal, VSCode extension, or an external app.
Line between editor/IDE is the former requires downloading extensions but the latter bundles them in the final application.
Debugging and code navigation are much better in proper IDE. In text editors you grep/search with regex.
Why on earth would you want to look at a code diff in a terminal? Why on earth would you want to use weird bespoke keyboard shortcuts to navigate sub-agents in a TUI session. Why on earth would i not want the ability to open/edit a file manually in the tool im using to write code.
can someone please explain this to me?
> Why on earth would i not want the ability to open/edit a file manually in the tool im using to write code.
I don't know what you mean by this. You can open any file in neovim at any time without leaving the program.
Being familiar with the terminal also makes building CI for the team trivial because I'm already familiar with how all the commands work in the CLI. I'm basically the goto 'devops' guy because I'm one of the few people who actually knows how to work in a Unix terminal.
I will say, TUI is not the same as CLI. I don't find a meaningful difference between a TUI and a GUI other than being able to use tmux or something for window management. I prefer gui tools for database management, querying, git diffs, email, all kinds of stuff.
As for the superiority complex. I've got no judgement on people who prefer the GUI. I have many excellent coworkers who primarily use GUI tools. Having said that, every engineer I've met that works primarily in the terminal has been great. It's a very strong signal of technical competence in my opinion, but terminal familiarity being a signal of competence, in no way makes GUI usage a signal of incompetence.
And that familiarity transfers between different systems. Windows, Mac, Linux, whatever: The flow of any particular terminal-based program is the same everywhere that it can be used.
It's tidy, and light. It's also network-transparent, and things like ssh keep it secure. Multi-user support is the norm instead of the exception on systems where terminals are common. It doesn't interrupt anyone else's work like something like using Anydesk to access some GUI desktop somewhere else can.
The keyboard shortcuts are annoying at first, but they're faster than mousing around in a GUI -- and once learned, they're approximately impossible to forget.
(You're free to hate terminals if you wish. I don't care if you justify it; I'm not your boss.)
- I work a lot with data - and streaming data through text tools is twitch fast. If someone has a question about data - before anybody else can log in to their superset, or analytics database, and try and work through the SQL queries or charts to get the answer - I've already jammed the data through awk and got an answer.
- As an SRE - I work with a lot of systems that have pretty rich APIs - so being able to send a request, get the answer back in json, dump it into jq, select the parts I care about - maybe -c to compress it and ripgrep a subset out - is just fast.
- I work in a lot of contexts with a lot of different systems, datacenters, applications - tmux lets me keep all of them cleanly organized in a separate windows and subpanes. I'll have 15-20 windows open per week, and maybe a 5-6 panes in each- keeping 100+ different contexts (and scroll backs, bash history) - all nicely organized is really useful.
- I'm also a systems guy - and there is no other way to dig into a system but the terminal - netstat, ps, dmesg, /proc - these are all components that have only one credible path to investigation and discovery. If you aren't super comfortable in the terminal - zero way to learn about this stuff.
- Working remotely - means ssh. So - once again - terminal.
The Focus on the terminal is that it's the best tool (and in some cases the only tool) for so many of these tasks - and by performing these tasks a lot - you learn about systems - so the people who spend a lot of time in the terminal tend to know a lot more about systems than people who don't.
So you can bring your agents to any remote system, it even works somewhat well for network devices.
If your workflow fits entirely within a single app's GUI, then yeah, the terminal version of that app is not going to be as useful. But if that app doesn't exist yet, you can put together an 80% version of it for 20% of the work.
Historically, it's also a lot more resistant to rot. Brian Kernighan isn't going to start charging a subscription fee for AWK - and if he did, there are many forks and similar tools.
And, addressing a specific point - why would I want to view a code diff in a terminal? Sure, 'diff old.txt new.txt' is probably less useful than popping it open in a nice GUI with highlighting. But "diff old.txt new.txt | grep '^+'" will only show me added lines, or "| less" and type "/foobar" to jump to all mentions of foobar.
And this is like, the least you can do - the stuff you learn in the second class of "Using the Terminal 101". You can easily use this with git, as a building block to make a quick script to graph the number of changes over time in your repo. Yes, there's probably a GUI somewhere in your workflow that can show this (maybe you click around in Github to find it). But, maybe you also want to just filter that to changes in a specific module in the codebase, or an author, or quantify what module changed the most each month. If you've learnt the building blocks, the scriptability of the terminal lets you put that together quickly.
I'm not arguing that these programs should not have a GUI, for that was the simplest way to use them, but the lack of command line functionality places a hard limit on the productivity of an organization, and ensures that the only progress on that front comes at the expensive of exceedingly limited developer time.
"But nobody knows how to use a terminal anymore," I hear you say. Well of course they don't, nobody under 35 without a background in programming has ever had cause to use one. We made everything so simple that nobody ever has to learn anything. That isn't to say that people cannot learn, just they have been robbed of the natural opportunity to do so. Otherwise intelligent people never progress beyond the manual step-by-step interaction that passes for "using a computer" today.
A computer is a tool in much the same way that the a machine shop is a tool: it is a tool that can build other tools. The role of software developers should not just be to build simple tools that can do one task in isolation, it should be to build tools that less technical people can use to build the tools they find themselves in need of. GUI-only programs are simply not fit for that purpose because they lack composability of simpler terminal-based programs.
Terminal/CLI is superior if you know what you want to accomplish. There is no faster way of doing stuff on a computer if you know what you’re doing and having tab completion and knowing all the incantations.
GUI is superior when you have to figure out what to do and what needs to be done. GIT branches state checking, finding out a switch to enable a feature so much easier in graphical interfaces.
I find that what you call „annoying engineers” are just people who have their tools and use them efficiently … but once they have to step out of their path they become annoying or even obnoxious.
The opposite is random person who expects everything to be easy and available in less than 3 clicks — well good luck engineering an interface for complex system that does that, not everything is Facebook, there are systems where you need to spend time wrapping your head around.
Once I got the tmux settings for proper scrolling and whatnot it feels fine. Honestly the TUI of tmux is the one that really enrages me - so much complexity for just "I want to switch terminals on my remote".
Why on earth would you want to look at a code diff in a terminal? Why on earth would you want to use weird bespoke keyboard shortcuts to navigate sub-agents in a TUI session.
can someone please explain this to me?
The terminal is an old but astonishingly powerful user interface that is still evolving.
Good terminals can be very snappy and configurable in ways that most GUI are not.
There is also arguably an aesthetic/fetishism appeal to it.
I've worked in the terminal at some point of my career, as there was not many other choices, and I understand how someone can get really used to it.
Why should others spend their valuable time helping you? Especially when you insult the people you want to answer you "fetish/superiority complex" just demonstrates your own prejudice.
Personally I ask AI for a summary of positions, and prompt to provide some good articles on a subject - ideally articles from supporters of either side.
To start a big part is just the efficacy of them, which comes down to the model and the harness logic itself. CC is good, it's sub-agents, loops, background jobs / agents, skills/hooks/etc have typically been pretty far ahead though others are constantly catching up.
But you're sort of missing something. I use iTerm, so to me it's not the TUI itself, it's iTerm. And while it's imperfect, what I get is this:
I can open and close sessions nearly instantly and tile and window and tab them as flexibly as I want, plus it's a system I'm familiar with in terms of shortcuts etc. Has my configured theme, fonts, etc all set up. Every GUI app is different, every TUI app has half of the UI already incredibly familiar to me, it's not "just text", it's iTerm.
That also means they all are the same - I run Codex and Claude and pi side by side, and i switch between them with no overhead and minimal mental model shift. Sure, different harness does suck but that's the same issue with GUI just with an additional new layer to learn.
Smaller thing is because it's all text, there's no limits on my ability to copy things out. And it's a really fast text renderer that can render tens of thousands of rows efficiently. Many GUIs have various dialogs, unselectable areas, virtualization, or just slow past a point. I trust my terminal scales.
Just a few reasons.
Anyways, I use cursor for a number of reasons:
1. I still want very quick access to the code in the editor. So I want the IDE.
2. Generally solid defaults. Auto-compaction, plan mode, etc, all work pretty well.
3. When I switched back to it from Claude code, it was genuinely faster at running Opus than Claude code. Claude code was grinding to a fucking halt every two minutes.
4. So annoying to search and view your chat history in Claude code. I’m a visual person. I also want all my repos loaded into a big workspace. Cursor also does that great out of the box.
5. I don’t have time to redo my terminal setup again to optimize it for Claude.
Tbh, I’m not aware of much that Claude code does that you can’t also do in cursor. At the end of the day, the agent loop and tools are not that different, and the model is identical.
The tool you use to prompt it is not the hard part. I just work faster when I have everything easily accessible in one spot, which was easier for me to accomplish with cursor than Claude. I found it just got out of my way.
Or it could be just Claude CLI doing something very well.
I don't think that will make much difference in a year.
Sophisticated chain of reasoning LLMs like ChatGPT have baked in some natural language operations and they make it so i can create at a higher level of the language expression stack. But I'm still formulating my own expression. There's no conceivable path I can see where an improved model is going to be able to do what I do. I think that is clear from my ChatGPT threads at least.
I think there's a pretty good chance that we've reached the point of diminishing returns, for our specific use case.
There are still like a billion other (more difficult) use cases to be tackled, but I think "generating code" has gotten really good to the point where the other bottlenecks will prevent further exponential progress on this specific task.
With deterministic workflows, type-safe languages and test suites, agentic loops pretty much “can’t fail”. They will continue until the types resolve, the tests pass, and the project requirements are deterministically met.
By that point it’s literally just a case of typing a prompt in to a text field, and waiting.
Similar experience, having transitioned from full-stack to a dedicated C/C++ stack, learned/experienced firsthand that there is no one-size-fits-all tool.
My one question is what popups exist in cursor? It is my daily driver and I cannot recall any popups.
If you do any kind of on-device work, it will spin up a local HTTP log server, and pipe logs from your real device (phone, hardware, etc.) to the server and do realtime debugging.
Claude will mostly guess, have you copy + paste logs, etc.
I’ve never used cursor and have only seen it in a couple work lunch and learn demos. I’ve never seen that feature. I have a lot of use cases where I’m asking cc to move a widget down a little bit or make a data table full width etc. Being able to reference the actual UI would be useful.
Tools like CC are best suited to vibe coding.
I actually don't let AI take on large tasks beyond test writing and refactoring helper scripts/utils. I keep it on a very short leash for driver/middleware code since the quality bar needs to be extremely high for our codebase. Up until recently I didn't even trust it for that, but some experiments show it's fairly good and even detected issues outside of the refactored functions which I did let it touch. This is with a good amount of 'thought engineering' though where I try to think hard about how to emphasize certain factors and define the problem as best I can.
to me, this seems like the perfect time for Cursor to exit and even "Q3 completion" is too late. Deal just needs to close. Fortunately Q3 completion could mean July 1st too
- Composer 2.5 is cheap, fast, and very effective; I don't even use other models that much anymore, as it's usually marginally better for way more cost, though sometimes I do for specific things like making better translations for our app than a coding-specific model could normally output
- It makes setting up and maintaining Cloud Agents super easy, the agent can basically set up itself and if anything changes that makes it not update properly, then it tells you and can fix itself easily
- Easy to move agents from remotely in the cloud to local and vice versa
- Easy to work with agents via either comments on GitHub, or via the web app on my phone (though it is relatively constrained relative to the actual desktop UI, which is a bummer)
- Code reviews with Bugbot is surprisingly good now vs when it first came out, as is the Security Agent, while being an order of magnitude cheaper than stuff like Claude reviews
- Automations are easy to configure and manage - crons, in response to repo events, etc. For example I don't use Renovate or Dependabot much anymore since LLMs can update deps and investigate subtle breaking changes much better than a dumb version bump script can
- Limits are obvious rather than Anthropic's mysterious quota amount that they don't explain at all
- Queueing up messages rather than the agent taking in new messages mid-work and then trying to mesh them together somehow - I find queueing much more predictable and easier to work with
- Plan mode is good too, but not particularly different than any other agent
- Easy to jump into the Editor view and actually go in and manually code things or interactively code with the LLM when you need to, since sometimes LLMs just suck at doing certain things autonomously. I'm not in the "stop coding bro" camp, I still like to take the wheel fairly often.
I'm not sure if it's because my personal projects are small enough I know them inside out and my work project is huge, but I prefer terminal code agents for day work over ide integration.
switched back to vscode so i'm not exposed to the potential mess that is openvsx too. trying to get used to zed but i'm just so used to vscode
How so?
That said, I think we're in a narrow window of time right now where any of this matters. Prompt "engineering" and working around your tools will be over in a year or so.
Fwiw I am a c/c++ systems engineer. I think anyone mentioning anecdotal experience like this should clarify. Maybe frontend JavaScript folks have a totally different take and that's expected.