61–70 of 518 posts
Earlier quoted context omitted.
I swear every time a new model is released it's great at first but then performance gets worse over time. I figured they were fine-tuning it to get rid of bad output which also nerfed the really good output. Now I'm wondering if they were quantizing it.
[flagged]
Despite the popular take that LLMs have no moat and are burning cash, I find OpenAI's situation really promising. Just yesterday, they reported an annualized revenue run rate of 10B. Their last funding round in March valued them at 300B. Despite losing 5B last year, they are growing really fast - 30x revenue with over 500M active users. It reminds me a lot of Uber in its earlier years—fast growth, heavy investment, b…
For OpenAI, the more people use the product, the same you spend on compute unless they can supplement it with another ways of generating revenue.
I dont unfortunately think OpenAI will be able to hit sustained profitability (see Netflix for another example)
Google has been catching up. Funny how fast this space is evolving. Just a few months ago, it was all about DeepSeek.
Earlier quoted context omitted.
Especially now that they are second in the race (behind Anthropic) and lot of free-to-download and free-to-use models are now starting to be viable competitors. Once new MacBooks and iPhones have enough memory onboard this is going to be a disaster for OpenAI and other providers.
OpenAI are second in the race to Anthropic in some benchmarks (maybe?), but OpenAI still dwarves Anthropic in distribution and popularity.
Despite the popular take that LLMs have no moat and are burning cash, I find OpenAI's situation really promising. Just yesterday, they reported an annualized revenue run rate of 10B. Their last funding round in March valued them at 300B. Despite losing 5B last year, they are growing really fast - 30x revenue with over 500M active users. It reminds me a lot of Uber in its earlier years—fast growth, heavy investment, b…
The problem is your costs also scale with revenue. Ideally you want to have control costs as you scale (the first you build is expensive, but as you make more your costs come down). For OpenAI, the more people use the product, the same you spend on compute unless they can supplement it with another ways of generating revenue. I dont unfortunately think OpenAI will be able to hit sustained profitability (see Netflix f…
Netflix has been profitable for over a decade though? They reported $8.7 billion in profit in 2024.
Earlier quoted context omitted.
I swear every time a new model is released it's great at first but then performance gets worse over time. I figured they were fine-tuning it to get rid of bad output which also nerfed the really good output. Now I'm wondering if they were quantizing it.
It seems that least Google is overselling their compute capacity. You pay monthly fee, but Gemini is completely jammed 5-6 hours when North America is working.
Personally I've found these bigger models (o3/Claude 4 Opus) to be disappointing for coding.
Opus is really great but through Claude Code. If you used Cursor or RooCode it could be normal that you get disappointed
Earlier quoted context omitted.
I swear every time a new model is released it's great at first but then performance gets worse over time. I figured they were fine-tuning it to get rid of bad output which also nerfed the really good output. Now I'm wondering if they were quantizing it.
I've heard lots of people say that, but no objective reproducible benchmarks confirm such a thing happening often. Could this simply be a case of novelty/excitement for a new model fading away as you learn more about its shortcomings?
Earlier quoted context omitted.
I swear every time a new model is released it's great at first but then performance gets worse over time. I figured they were fine-tuning it to get rid of bad output which also nerfed the really good output. Now I'm wondering if they were quantizing it.
I've heard lots of people say that, but no objective reproducible benchmarks confirm such a thing happening often. Could this simply be a case of novelty/excitement for a new model fading away as you learn more about its shortcomings?
Anecdotally, it's quite clear that some models are throttled during the day (eg Claude sometimes falls back to "concise mode" - with and without a warning on the app).
You can tell if you're using Windsurf/Cursor too - there are times of the day where the models constantly fail to do tool calling, and other times they "just work" (for the same query).
Finally, there's cases where it was confirmed by the company, like Gpt-4o's sycopanth tirade that very clearly impacted its output (https://openai.com/index/sycophancy-in-gpt-4o/)
It's going to be a race to the bottom, they have no moat.