extr

12 ms·

181.
by extr 1y ago
I think this is an interesting and cool direction for Cursor to be going in and I don't doubt something like this is the future. But I have my doubts whether it will save them in the short/medium term: - AI is not good enough yet
182.
by extr 1y ago
"The same editor you already use for free, but with slightly nicer UI for some AI stuff" is not a $900M ARR product.
183.
by extr 1y ago
Cursor's tab completion model is legitimately fantastic and for many people is worth the entire $20 subscription. Lint fixes or syntax-level refactors are guessed and executed instantly with TAB with close to 100% accuracy. This is the
184.
by extr 1y ago
Not sure what you mean. "Hosted agents in sandboxed/imaged environments"? The entire selling point of CC is that you can do - > curl -fsSL http://claude.ai/install.sh | bash - > claude - > OAuth to y
185.
by extr 1y ago
IIRC problem is that VS Code does not allow extensions to create custom UI in the panels areas except for WebViews(?). It makes for not a great experience. Plus Cursor does a lot with background indexing to make their tab completion model r
186.
by extr 1y ago
IMO other than the Microsoft IP issue, I think the biggest thing that has shifted since this acquisition was first in the works is Claude Code has absolutely exploded. Forking an IDE and all the expense that comes with that feels like a was
187.
by extr 1y ago
This is the killer app for LLMs for me. I used to get super bogged down in the details of what I was trying to do, I would go a whole afternoon and while I would have started on the feature - I wouldn't have much to show for it in term
188.
by extr 1y ago
I find Gary's arguments increasingly semantic and unconvincing. He lists several examples of how LLMs "fail to build a world model", but his definition of "world model" is an informal hand-wave ("a computationa
189.
by extr 1y ago
The author seems to be under the impression that AI is some kind of new invention that has now "arrived" and we need to "learn to work with". The old world is over. "Guaranteeing patches are written by hand" is
190.
by extr 1y ago
AI slop
191.
by extr 1y ago
Really interesting, I know nothing about cinematography and don't really have the language to articulate what it is about their clips that are superior to other AI visuals, but it's definitely noticable, even if the overall qualit
192.
by extr 1y ago
We use it by having a Large Model delegate to Flash 2.0. Let's say you have a big collection of objects and a SOTA model identifies the need to edit some properties of one of them. Rather than have the Large Model perform a tool call o
193.
by extr 1y ago
Gemini Flash 2.0 is an absolute workhorse of a model at extremely low cost. It's obviously not going to measure up to frontier models in terms of intelligence but the combination of low cost, extreme speed, and highly reliable structur
194.
by extr 1y ago
I've had an identical experience of ChatGPT misidentifying my kid's rash. In my case I would say it got points for being in the same ballpark - it guessed HFM, the real answer was "an unnamed similar-ish virus to HFM but not
195.
by extr 1y ago
Modeling the distribution that produced a piece of text is what LLMs literally exist for, so in some sense this is unsurprising. But it calls into question almost all existing alignment research.
196.
by extr 1y ago
To answer your question directly "it depends". There are a lot of different subtasks to coding. Defining requirements, sketching/discussing architecture, coming up with a technical plan, executing it, creating tests, etc. Dif
197.
by extr 1y ago
If you live in a metro area it's really not like that. Prices are higher but my credit card gives me free Uber One, so fees + tip are managable. Plus Uber/Postmates offer (at least for me) extremely aggressive coupons pretty often
198.
by extr 1y ago
It's possible but I think this is a bit of a non-issue for cursor. Microsoft extensions are pretty good but are not irreplacable, and in the meantime cursor has grown astronomically fast and has grabbed a ton of "AI Coding" m
199.
by extr 1y ago
I use cursor primarily because of the great tab autocomplete model, but I've always thought it was a bit scummy they blatantly violate the VS Code licensing. Windsurf ships a special version of the pyright extension for this reason. Wh
200.
by extr 1y ago
Yeah this is a great summary of what I do as well and I find it very effective. I think of hands-off AI coding like you're directing a movie. You have a rough image of what "good" looks like in your head, and you're tryi
201.
by extr 2y ago
Didn't know they did the PB&J thing at Harvard. I remember doing that in the 3rd grade or thereabouts.
202.
by extr 2y ago
The problem I find in many cases is that people are restrained by their imagination of what's possible, so they target existing workflows for AI. But existing workflows exist for a reason: someone already wanted to do that, and there h
203.
by extr 2y ago
No idea, I believe it's just a fixed point. Personally I use it while sitting in front of my Mac about 1-2 feet from my face. I've done tests, it's better than every other form of audio input I have available, including stand
204.
by extr 2y ago
Biggest thing is you need a nice mic that's very close to your face, like you might see on a twitch stream. Good noise isolation via a directional mic off-camera is quite difficult/expensive apparently.
205.
by extr 2y ago
It's not just good, I found it to be way better than a standalone shotgun mic connected via USB. I researched this for WFH and found a lot of people saying you were going to spend hundreds to replicate the quality in a more "profe
206.
by extr 2y ago
Eh, so in reality there are a lot of AI products people are trying to build and it's very unclear at the outset "if it's possible", where "possible" is a business question that includes factors like: - How hard
207.
by extr 2y ago
I bring this up often at work. There is more ROI in assuming models will continue to improve, and planning/engineering with that future in mind, rather than using a worse model and spending a lot of dev time shoring up it's weakne
208.
by extr 2y ago
The fundamental reason is that RAG kind of sucks and requires a ton of effort/optimization to reach a high degree of reliability for most applications. RAG solutions are not Anthropic's core product. Just reading all the relevant
209.
by extr 2y ago
I think in my head I just assumed "assembly line" = hard labor, like a sweatshop. Your dad's version of "assembly line" sounds like exactly what I'm talking about. I just want to keep the machines running.
210.
by extr 2y ago
Okay, well whatever the job is called where you go out on the floor and tweak the gizmos. I'll do that if I have to.

More ›