afro88

18 ms·

181.
by afro88 1y ago
If Qodo are reading this: please introduce a plan that isn't for teams or enterprise. A "pro" plan for individuals who want more than 250 credits per month.
182.
by afro88 1y ago
Claude 4 is definitely warmer and more empathetic than other models, and is very reliable (relative to other models). That's a huge counterpoint to this paper.
183.
by afro88 1y ago
I have a real world problem I gave o1 when it came out and it got it quite wrong. It's a scheduling problem with 4 different constraints that vary each day, and success criteria that need to be fulfilled over the whole week. GPT-5 Thi
184.
by afro88 1y ago
Did you use the thinking model? I tried it twice and got a similar result to the author each time. "Thunder rolls through the page. Here, words bend reality. I can compress wild ideas into whispers. See how tightly they fit. In one swe
185.
by afro88 1y ago
First question: https://chatgpt.com/s/t_6899bc7881e88191bb3d2146eac718d7 If you meant, after you converse with it for a while: what was the conversation leading up to this point?
186.
by afro88 1y ago
Does anyone ever get answers this short? What's the system prompt here? That may bias things a little. Also it's GPT not GTP
187.
by afro88 1y ago
I tried it again today out of curiosity. OpenAI said there was some routing bug on launch and requests were going to the cheaper model. Today it seems pretty good. Not perfect, but not a spectacular failure. https://chatgpt.com&#
188.
by afro88 1y ago
Claude Code but can use GPT-5 built in. Not a bad selling point
189.
by afro88 1y ago
In my experience Opus 4 isn't as good for day to day coding tasks as Sonnet 4. It's better as a planner
190.
by afro88 1y ago
OP is asking for where to get started with Claude for coding. They're confused. They just want to mess around with it in VSCode. And you start talking about Ollama, PAT, coding your own wrapper, composing a system prompt etc.!?
191.
by afro88 1y ago
You want all that on a resume?
192.
by afro88 1y ago
The filter is actually "site:", but yes, no links indexed on DDG
193.
by afro88 1y ago
Being able to do big refactors quickly in the moment really helps in a solo dev environment, but in a team it puts a lot of review (and QA) burden on them. It makes me wonder if we're moving towards a teams model where individuals own
194.
by afro88 1y ago
A HUD is an even more "confident" display of data than text though. What do you do with a HUD that hallucinates? Is there a button on each element that shows you sources?
195.
by afro88 1y ago
It's crazy to me that someone can write a post called "How We Rooted Copilot" when in reality they got root in an ephemeral python sandbox container that was locked down so much that they couldn't do anything. I read &qu
196.
by afro88 1y ago
> In internal evals we improved Cursor’s performance by 27 % once Nia had indexed external docs models couldn’t get from their training data or searching the web. What external docs do you have access to that aren't found on the web
197.
by afro88 1y ago
Sigh, there's that study again, quoted without any "early 2025" context. In early 2025 AI wasn't very effective, especially when used by engineers that didn't have an intuition for what to and what not to use it for
198.
by afro88 1y ago
Source? That's genuinely surprising as someone living outside the US
199.
by afro88 1y ago
> Error rates compound exponentially in multi-step workflows. 95% reliability per step = 36% success over 20 steps. Production needs 99.9%+. This misses a key feature of agents though. They get feedback from linters, build logs, test run
200.
by afro88 1y ago
Well documented helps a lot too. You can use an LLM to help document a codebase, but it's still an arduous task because you do need to review and fix up the generated docs. It will make, sometimes glaring sometimes subtle, mistakes. An
201.
by afro88 1y ago
This is almost the opposite of what OP is asking, and what the post from antirez describes.
202.
by afro88 1y ago
You can actually just put Cursor in manual mode and it's the same thing. You 100% manage the context and there's no agentic loop. If your codebase fits in the context window, you can also just turn on "MAX" mode and it p
203.
by afro88 1y ago
You moved the goalposts from "do something catastrophic by doing something completely unrelated to what you were trying to do" to "press the wrong pedal or button in a car or elsewhere" We're talking about taking an
204.
by afro88 1y ago
I said this when the linked paper was shared and got downvotes: it's based on early 2025 data. My point isn't that it should be completely up to date, but that how we need to consider it in that context. This is pre Claude 4, Clau
205.
by afro88 1y ago
I pull my handbrake every time I park my car, but never mistake it for the windshield wiper while the car is moving
206.
by afro88 1y ago
Early 2025. I imagine the results would be quite different with mid 2025 models and tools.
207.
by afro88 1y ago
Some great info, but I have to disagree with this: > Q: How much time should I spend on model selection? > Many developers fixate on model selection as the primary way to improve their LLM applications. Start with error analysis to un
208.
by afro88 1y ago
This is a strawman isn't it? I haven't read one post or comment saying that writing code is "the bottleneck". It's something that takes time. That time is now greatly reduced. So you can try more ideas and explore p
209.
by afro88 1y ago
It coming from computer science might be the issue. There's a lot of open source repos out there that have tricky bugs, and todo lists of features that are too complex or time consuming for casual contributors to tackle. Adding signifi
210.
by afro88 1y ago
This is great. Thanks for creating it! Matches how I read comments, and fixes my only gripe with HN: I'm forever missing the touch target of the little [-] to collapse a thread once I have read it. Yes there is a next button too, but I

More ›