We're truly building walls everywhere.
Personally, I tried copilot when I got it for free as a student and it didnt make a difference. The reason I know is that I was coding on two devices, one which had copilot installed the other didnt, and I didnt care enough to install it on the latter through an entire semester.
Its just slightly better autocomplete, by a questionable standard of "better".
There’s literally nothing an llm can write or tell you that you can’t write yourself or find in a manual somewhere.
That's like saying, there's literally nothing a service business can do for you that you can't do yourself. It's only true in a theoretical sense, if neither time nor resources are a constraint.
In such hypothetical universe, you don't need a dentist - you only need to spend 5+ years in medical school + whatever extra it takes to become proficient with tools dentists use + whatever money it takes to buy that equipment. You also don't need accountants, lawyers, hairdressers, or construction companies. You can always learn this stuff and do it yourself better!
Truth is, time and attention is finite. Meanwhile, SOTA LLMs are cheap as dirt, they can do pretty much anything that involves text, and do it at the level of a mediocre specialist - i.e. they're literally better than you at anything except the few things you happen to be experienced in. Not perfect, by no means error-free - just better than you. I feel this still hasn't sunk in for most people.
Also, local llms with an agentic tool can be a lot of fun to quickly prototype things. Quality can be hit or miss.
Hopefully the work trickles down to local models long-term.
I think you are approaching this with the wrong mindset. I see it as I'm paying somebody to type and document for me. If you treat LLMs like a power tool, it is very easy to do a cost benefit analysis.
So we're going back to the last century, but given we are in a different computing context, only the stuff that can be gated via digital stores, or Web Services, gets to have a way to force people to pay.
But I am glad we now have more paid options available. Tooling is important and people that do good work should be able to charge for high quality tools.
I would be much happier in a world full of tools licensed like Sublime Text, where I can purchase a license and just run it without the need to constantly phone home though.
Nothing stopping you to build the world you want really.
There's no moat, all the clever prompting tricks Cursor et al. are just that - there is no secret sauce besides the model at the other end.
Complexity isn't an issue either, have the model write the interface to itself.
I'm not understanding what it is about a private company launching a product that changes that?
You can do it without IDEs, nothing is stopping you. I don't think this is a new phenomenon though.
You are free to coding without spend a dime, these AI dev tool cost money because these LLM cost money to run
You can get the same experience with open source tools that you can run your own model on your pc
I mean you don't need to if you don't want to. I am gainfully employed as a software developer and what I do everyday is literally just fire up Emacs on my Linux machine and write code. To this day I haven't figured out what llms are supposed to do that a bunch of yasnippets don't.
Just like five years ago most of my day is reading and debugging code, I'm not limited by how fast I can type.
It is kind of terrifying that I probably would stop coding for the day if those subscriptions end. (I get far too much convenience out of them)
I have tried to rationalize it by the fact that I do pay for internet, and version control, and my peripherals etc
The problem is that coding was a passion, but turned out to be very lucrative profession so loads of people who can't do it want to do it.
This is why we have languages like Go, and AI tools: allow people who don't want to learn how to be developers, to get a job as developers.
Also 20$ per month is way less than what it costs them to run it. Eventually they will need to charge way more to cover their costs, and the people who can't code without an AI assistant will need to pony up :)
https://github.com/kelseyhightower/nocode
Not even a commit: https://github.com/timdorr/-/commit/9e5a571abd3fc4f8714e8c40...
This is what a commit-less repo looks like on GH: https://github.com/verdverm/_
Unfortunately they don't show the description "starting repository for any language and project" (a play on the _ (any) token in CUE)
But if it's for some personal project that you're putting on Github, does it matter? If my code is going on Github anyway, it's going to get slurped up regardless. I don't particularly care if it gets processed by Cursor or ByteDance before it gets scraped.
I don't think the fear is that they'll steal code you'll end putting publicly on GitHub, but everything else. I guess there is some fear that it won't just analyze and process what you currently have open, but might scrape your computer for more data and so on.
I personally don't believe ByteDance would be stupid enough to even attempt exfiltrating files from developers machines, which typically are better protected than the average user computer, just trying to see the perspective of others with a more charitable reading :)
Correct.
I laugh a little every time I hear some folks immediately push the “Big Bad China is just trying to steal our data” narrative. My immediate reaction is, “Grab some tea and let’s sit down real quick, so I can tell you what some of our companies right here in America and our government do with our data.”
If that happens to be an editor from China instead of US, I don't know what difference that would make? Both governments in those countries are crazy about spying on both their own citizens and everyone else, and have their corporations under surveillance.
As long as the editor doesn't send opened files/files from my drive to some remote backend, I couldn't care less about the nationality of the developers.
I see your point but in this particular case, I think it needs to send the opened files to a backend for processing
Is this sarcasm? Isn’t that exactly what the editor has to do to do its job? How else would the AI stuff work? It’s not running on-device, right?
(I am guessing that there will be a deal floated for Elon to buy TikTok in a few weeks)
Although, if TikTok's earnings can cover the payments on the loan, it could be possible.
Apparently, the executive order was illegal
It was banned because of a law passed by Congress. Representatives aren't bureaucrats.
They have a lot of companies with very large consumer, and their market is saturated already. Time to expand to the rest of the world.
That being said, I am using (Chinese) Deepseek r1 because there isn't currently a free LLM on par with it. I am careful with what I share though, a little more so than with any others that are not locally fun.
So as a non USA citizen, I'm way more afraid of USA than I am of China.
Which is why I would like the US govt to have as little data as possible about me (who knows what the mainstream politics will look like, things that are very innocent today might be punishable by death in 10 years).
Chinese govt, on the other hand, can have my data freely, since I don't think I'll ever move to the Mainland China.
For example, China killed a lot of people. They were all Chinese.
Look at conquered regions like Tibet and Xinjiang.
Look at how they set up police forces in foreign countries to keep an eye on Chinese citizens living abroad. Even having kidnapped and illegally held Chinese citizens in England because they posted anti-CCP messages on WeChat.
Look at countries like the Phillipines (not even a direct neighbour) who are trying to hold on to small fishing islands just off their coast because the CCP claimed those islands in the 1970s.
Remember a few years ago when they ran a week long military exercise around Taiwan... because a US representative spoke to the Taiwanese president. Sure I agree with you that the US also overstepped a line here, but for your response to be shelling the waters around the island is excessive to say the least.
Look at the aggressive nationalist and imperialist news they feed their own population, and the propaganda spread to make the Japanese seem like demons. Did you know there are several theme parks in China where children are encouraged to Bayonette a mannequin of Japanese Imperialist soldiers?
Call the US as bad as you like, but I've never seen a theme park where children are actively taught how to kill and demonise the Taliban or Nazis.
[Edit: hey Europeans commenting and downvoting below, note the words "in principle" in the above comment and evaluate which of the two countries do or do not purport to stand for these things despite whatever your hot take may be on the current moment.]
Have you tried the 7b?
Employees of Bytedance (and other Chinese companies) have to deal with draconian employee rules and agreements that tie them back to the rules of their Chinese mainland parent corporation and the Chinese government itself. For example look at the details that came out in this lawsuit against TikTok, where employees have to agree to uphold Chinese national interests, uphold socialism, etc.
https://dailycaller.com/2025/01/14/tiktok-forced-staff-oaths...
Early TikTok was basically a standard Chinese startup in terms of how they collected data to optimize the app... It's not because they cared about what "you" are doing.
The Chinese tech ecosystem is just more competitive than the US one.
Most social media companies have been careless with data until they got public backlash though... TikTok's systems are probably more secure today than anyone else's... Because of the backlash.
Does that sound like a business or an intelligence front?
Tencent's Hunyuan 3D - Tripo, etc.
Tencent's Hunyuan Video - Sora, Runway, Pika, etc.
ByteDance's Trae - Cursor, etc.
DeepSeek R1 - GPT, etc.
Unitree - Figure, Tesla, Boston Dynamics, etc.
And a ton of this stuff is completely open source. All of the Tencent stuff is. This destroys the moat of so many companies that have spent hundreds of millions of dollars training and building their tech. It's just out there for free for anyone to build with.
And has anyone been checking the volume of Chinese AI research papers? Almost every impactful paper I've read in the last month has mostly Chinese names. And a lot of those papers come with code. Usually permissively licensed.
China is absolutely killing at the AI game. They've fast followed (or in many cases led) into a position of strength.
I wouldn't want to be RunwayML, Pika, Luma, or Tripo right now. Any "foundation model" company with a single use case is getting cloned and commoditized.
edit: Please don't downvote me because you don't like the message. I'm actually fine with you shooting the messenger, but this is absolutely worth talking about. It's pretty surreal to watch this all start to unfold over the last quarter or so. I want to read what others think about it.
It is that bad that all the American companies withrew from the public benchmark library. You can only see there Chinese commercial solvers and open source ones.
But in my opinion this was completely expected, most research in this area is now done in China, American universities have in comparison stopped in time.
Love this field of CS!
(I am not in AI, so please correct me if this is inaccurate.)
(Source: frustrated friends who lose the government contract to someone with inferior technology)
I just had to talk someone I know out of using RedNote as the TikTok shutdown loomed last week -- they had ZERO IDEA about its provenance. Unless you're paying attention, it's actually not clear what's what.
How many products keep sending all kind of unknown telemetry, have a ton of trackers, etc?
Can't really blame them doing the same.
1) They try similar like Zuck with LLama make sure GPT or Claude or Gemini is not the real winner in the West world like Android/iOS was for smartphone, Facebook/Whatsapp for social media/IM, Google Search for search, Windows for desktop OS. Majority people get used to to single product and they don't switch often.
2) Trying to kickstart community similar like LLama kickstarted big community around it that helped with tooling, testing, etc.
3) Slowing down development of ai companies in the west - giving them less data for training (after all midjourney, elevenlabs, llama kickstarted training on copyrighted content, it's pretty sure they using user data at least from e.g. free version of GPTo-mini) and bleeding their budget kind of war of attrition.
4) Familiarise the west audience with their ai models - after all still not many people using models like DeepkSeek since not many providers host them and majority of people don't have fast enough computer to run full models fast enough - Trae helps with that
Or are you being sarcastic? You must be being sarcastic.
If you really want an integrated experience, and not just a sidebar UI, you need to go the same route as Cursor and fork Code-OSS (the MIT-licensed part of VS Code, analogous to Chromium for Chrome)
I don't think it has any special api access?
Wondering if it is a fork of vscode though because right now it will be competing against cursor.
I've also blankly asked it if there's any issues with the current code, and found errors before I try to build / run a project.
But the moment it gets a little less basic, its a different story. I absolutely hate the experience. So I love it and I hate it at the same time. Is it going to replace me? I don't feel like its there yet. And this has been the case for a while now.
There is just too much money and people invested in this that its hard to say anything negative about it. Nearly every VC has rebranded their websites around an AI-driven future. And, to be fair, it’s not a fad—it genuinely works well for many things. But for now, I’m still skeptical about how far it can really go.
I'm terrified to install a binary like that
Use something like emacs/vim/kate/eclipse/qtcreator if you want to avoid "phone home" software.
If a Chinese company does the same, at most someone in China will shrug.
Does anybody feel the same? I feel like nobody touched upon this, but it's always very nice to feel interfaces are responsive.
I accessed this through a Qubes AppVM (no GPU, limited memory and CPU budget) and the presence of videos makes this a very slow scrolling experience for me.
In general, anything that involves JS/CSS animation/blur/effects makes sites pretty slow (up to unusable for me). The unlogged homepage for github.com for example, spins my cpu at 100%.
But I guess instead of writing great libraries to avoid repeating ourselves, we drop all code to a lake monster which eats it and spits out answers.
Thus it supporting English and Chinese languages.
And it may be a competitive step taken after Alibaba launched their new coding assistant.
Although I haven't tried to find a Chinese mode myself.
That it's targeted at Chinese abroad is just what I saw from Googling it and reading the SCMP (Singapore) article on it.
The company that made it is also from Singapore (even if it's under Bytedance's umbrella).
They seem to be separate from TikTok. The only thing they have in the app store for example, is some chatbot application.
¹: which is dozens of them at this point, open and not.
²: with lots of speakers and training data.
Regardless, what is this exactly?
Also, anybody knows a decent IDE where AI is a first class citizen? So far my experience with plugins is, that they have too little context to be actually useful.
Sidenote: That marketing video was nauseating, it was moving too fast and didn't show any features for long enough or for enough steps.
If I ever have a need for buggy crap code, any random boot camp dev can write it for no money at all.
Makes you think about how they're extracting value from you to make up for this. I rack up non-significant amounts when using Claude 3.5 Sonnet for coding.
Considering the data they siphon from their other "free" apps, I don't want to think about what this does on developer machines with code bases and production access.
Recommended.
(But of course it would be nice to be able to run it against a local model which doesn't seem possible at the moment.)
- Void
- Melty
- Pear
- Aide
Just another AI wrapper around VS Code smh
Effort would have been much better spent on getting AI auto-complete "merged" with LSP. But who is listening? Pretty much everyone these days think they can plug an LLM and have their life issues sorted out.
It's sad because I have seen some truly remarkable progress in LLMs, but I feel like we aren't allowed to be honest anymore that LLMs aren't going to replace programmers or moderate our expectations.
For any semi novel development I want to write it myself to establish the patterns I like and get an understanding of what I'm building. It's the boring stuff I hand over to Claude.
I certainly can see it being useful for writing additional tests, although I am very worried based on what I've seen that it will introduce more bugs than it's worth, which is why it's still a tool I use for hobby projects and not for real work.
But literally every single programmer I know has this opinion.
With what I was left with at that point, it probably would have been a wash to fix that code or just start from scratch manually.
I don't blame Loveable, I think it's a wonderful product bordering on magic, like v0 and Bolt.new. I just think they are amazing products that have been over hyped to god-like status.
What it is just a summary of the internet.
I used to think he was competent but now I think he's a moron
Doing basic stuff with a very limited scope that’s likely well documented all over the internet, great.
Boilerplate, tedious yet simple things, works awesome as enhanced autocomplete.
More complex stuff, give it a shot, maybe you’ll get lucky, otherwise if it starts fucking up I’ve had the most success just taking a step back and doing it myself, maybe chatting with it as a live docs substitute, or a realtime stack overflow/discord programming channel which are also sometimes of dubious quality but frequently useful.
The misery really lies in getting stuck in that cycle of it just messing up over and over as you try to get it to make this thing work that is clearly beyond its scope and it’s just turning everything into a greater and greater mess of hallucinated bullshit.
I've built a couple pretty simple apps that have been basically no code beyond tweaks for me.
I do think a big knowledge gap is the acceleration of prompting by knowing how to talk about code. LLMs show a lot of difference in response between "make the top bit stay on the screen on small screens" and "make the <th> element sticky for viewports <600px".
- autocompletion (about half the time)
- questions on the syntax for certain commands (faster than looking up in the docs, usually pretty accurate)
- questions for how to solve a particular problem or gotcha (faster than looking up on SO, but also less accurate)
It's a useful tool for certain cases, for sure. But it's not a "game changer" in terms of productivity.
So I definitely think the value proposition of AI coding is higher when you have programmatical control over the workflow.
On a related note, the recent Cline Plan+Act function has also been a game changer.
LLMs are a tool. It’s also weird and counterintuitive so it takes some time to get really good with them.
I haven't heard anyone in tech say programmers will be replaced by this tech.
I have heard persons outside of tech say it, but I think they lack sufficient context.
I mean, if we get AGI, yes, we might be fully out of a job. But, so far I think many programmers agree this seems to be a ways off.
But, I think you underestimate how many programmers used to copy paste code from StackOverflow and articles, etc.
We even had/have a phrase for that: "copy pasta".
Those programmers are getting more applicable templated code than they used to get via copy paste.
When I'm coding in my preferred languages, I am faster without AI.
But, when I'm writing yaml or something else for an unfamiliar tool or platform, I do get a productivity boost, even if I have to debug the code / configuration.
Also don't believe your lying eyes.
- vs code
- vs
- IntelliJ IDEA
- Notepad++Everything that has been coming out of China is better than anything our mega trillion dollar corporation are producing and they are doing with with sanctions on the best GPUs.
I don't work in big tech, why are they so bad with sigifically more money and more advanced technology?
Google had a decade and a half head start and they can't even produce a model that is as good and as cheap as Deepseek building with no compute compared to Google.
I have been using Trae and VS Code with Copilot is laughable compare too it despite using the same models. So we can't even prompt LLMs as good as they do and they don't even speak English as a first language LMFAO.
We might be fucked and this is just the beginning.
Maybe time to learn code should be time to learn Chinese.
At the forefront of our politicians’ minds are how many genders there are. In the UK just building a new high speed railway has practically been the Manhattan project. Will we get back on track? I think the chances of that are rapidly diminishing.
Only problem is now I have no idea what they are talking about half the time.
Some C-levels newphew into computers and need a school project or something?