Yusefmosiah

Yes, I experienced this too. Is it a model issue or something related to the configuration and system prompts of Cursor?

It’s not just the quantity of tokens in context that matters, but the coherence of the concepts in the context.

Many conflicting ideas are harder for models to follow than one large unified idea.

To throw shade without directly showing where Wolfram goes wrong is pathetic. This blog reads like a post about basketball in the 90s that doesn't mention Michael Jordan, titled "Bull in a China Shop".

Thanks, but that comparison is for old models, a different, non-shipped version of Devin called “Devin-base”, and doesn’t include Claude.

Slack integration, automatically pushing to CI, etc., are relatively low-value compared to the questions of “does it write better code than alternatives?”, “can I depend on it to solve hard problems?”, “will I still need a Cursor and/or ChatGPT Pro subscription to debug Devin’s mistakes?”

Looking for comprehensive benchmarks with Devin vs Cursor + Claude 3.6 vs ChatGPT o1 Pro.

In my own experience using Cursor with Claude 3.5 Sonnet (new) and o1-preview, Claude is sufficient for most things, but there are times when Claude gets stumped. Invariably that means I asked it to do too much. But sometimes, maybe 10-20% of the time, o1-preview is able to do what Claude couldn’t.

I haven’t signed up for o1 Pro because going from Cursor to copy/pasting from ChatGPT is a big DevX downgrade. But from what I’ve heard o1 Pro can solve harder coding problems that would stump Claude or o1-preview.

My solution is just to split the problem into smaller chunks that make it tractable for Claude. I assume this is what Devin’s doing. Or is Devin using custom models or an early version of the o1 (full or pro) API?

The reason it makes sense to compare them is there are problems that Claude 3.5 (or o1) can’t solve. Can Devin solve them? If yes, it’s easily worth the $500. If no, it’s a harder sell.

This has a lot of potential. I've been using Cursor Composer heavily and it's great but buggy, and could be more agentic.

After about an hour with Windsurf, I find myself frustrated with how it deals with context. If you add a directory to your Cascade, it's reluctant to actually read all the files in the directory.

I understand that they don't want to pay for a ton of long-context queries, but please, let users control the context, and pass the costs to the user.

It's very annoying to have the LLM try to create a file that already exists, it just didn't know about it.

Also, comments on the terminal management reflect a real issue. One solution is to expose the Cascade terminal to the user, letting the user configure the terminal in a working state, so that it has access to the correct dependencies and the PATH is properly sourced.

Much of the criticism comes with a tinge of envy. "How did those guys get funding and not me?"

Envy is insidious. Focusing on how other people have unfair advantages eats away at your principles, motivation, creativity, and presence of mind to capture the opportunities available to you.

As a founder myself, I appreciate the fact that it's so easy (for some people) to raise money. It means that the competition is softer than it seems.

OpenAI’s strategy has been bizarre since at least last November, when they launched custom GPTs, then had the boardroom coup.

Since the launch of Claude 3 Opus, and then Claude 3.5 Sonnet, they have been significantly behind Anthropic in terms of the general intelligence of their models. And instead of deploying something on par or better, they are making demos of video generation (Sora) or audio-to-audio models, not releasing anything.

GPT-4o is quite bad at coding, often getting stuck in a loop, and “fixing” buggy code by rewriting it without any changes.

GPT-4o is speculated to be a distillation of a larger model, and now GPT-4o-mini is an even dumber smaller model. But what’s the point?

Who is actually using small/fast/cheap/dumb models in production apps? Most real apps require higher reliability than even the biggest/slowest/priciest/smartest models can provide today. For the use case of transformers that has taken off, aiding students and knowledge workers in one-off tasks like writing code and prose, most users want smarter, more reliable outputs, even at the expense of speed and cost.

GPT-4o-mini seems like a move to increase margins, not make customers happier. That, like demoing products without launching them, is what big old slow corporations do, not how world-leading startups operate.

When your CEO turns the brand from a status symbol to the opposite, it's "concerning".

Intergenerational conflict is particularly salient for wealthy Americans. The death of the boomers enables the millennials’ inheritance. Millennials blame boomers for not gracefully ceding power in their own lifetimes. But boomers see millennials as waiting for their parents to die, as complacent and heartless.

This conflict does not exist to nearly the same degree in families without wealth to inherit, and in cultures that respect the wisdom of one’s elders; rejection of one’s elders/youth worship is one value that millennial Americans inherited from Baby Boomers.

As to the emails, Clegg was right when he said this:

> For Millennials, authenticity, agency, autonomy, idealism, altruism etc all seem to be top of their list of desirable virtues – for better or worse, they are difficult to reconcile with Silicon Valley these days.

Well, it’s worse, not better.

Anyway, this kind of self-conscious persona-construction is opposite to what millennials vibe with. No matter how much Zuck styles himself like a rapper or athlete, he only appeals to boomer or at best Gen X sensibilities. This is great for META stock price, but doesn’t make Zuck appealing to millennials.

Replying to signal boost this. I too get hair loss from DHT-boosters such as creatine or the herb tribulus terrestris. I should note that I have a higher than average amount of body and facial hair, so likely have naturally elevated DHT levels.

When I stopped taking creatine, I did lose a few reps in all my lifts; it does work.