2001zhaozhao

8 ms·

91.
by 2001zhaozhao 3mo ago
Translation: other labs have learned to benchmaxx SWE-Bench Pro better than they do
92.
by 2001zhaozhao 3mo ago
Harness-wrapper tools that support multiple harnesses and allow sharing workspace features (skills, slash commands, etc.) between them will be meta.
93.
by 2001zhaozhao 3mo ago
Around Opus 4.7 level would be the same as Sonnet 5 while being cheaper overall. I wonder how good their subscription discount is on both their subscription types.
94.
by 2001zhaozhao 3mo ago
Just give it a scifi robot voice and pretend you're on a spaceship
95.
by 2001zhaozhao 3mo ago
Wait Devin has a CLI? Time to support it in my agent IDE just like Cursor's...
96.
by 2001zhaozhao 4mo ago
Image generation built for your very own dystopia!
97.
by 2001zhaozhao 4mo ago
Sonnet has crazy high usage limits on claude right now but it'll get 50% more expensive after July ends.
98.
by 2001zhaozhao 4mo ago
At least subscription users only have to pay $700 for $1000 of extra credits.
99.
by 2001zhaozhao 4mo ago
There are two wrinkles to this: - For Claude.ai subscriptions I think Sonnet is much cheaper than Opus. This is why there was a "Sonnet only" usage bar for Max tier for the longest time. - For some tasks the sheer amount of raw in
100.
by 2001zhaozhao 4mo ago
They are obviously trying to avoid getting Sonnet 5 blocked.
101.
by 2001zhaozhao 4mo ago
The $1/year numerical .xyz domain is pretty affordable already, and there are multiple providers now with free DNS services.
102.
by 2001zhaozhao 4mo ago
If so, it's impressive they managed to benchmaxx Qwen even further than it's already benchmaxxed.
103.
by 2001zhaozhao 4mo ago
+1, this is a good thesis. It's plausible and if it's true, then that explains the opaqueness of this whole ordeal and why it's been applied to both OpenAI and Anthropic models now.
104.
by 2001zhaozhao 4mo ago
We need people to rediscover remote development desperately, lol.
105.
by 2001zhaozhao 4mo ago
> game servers that run user-supplied scripts I would be incredibly surprised if these micro VMs are actually viable for UGC scripts in game servers without completely destroying one's wallet.
106.
by 2001zhaozhao 4mo ago
> back tracking constraint solver We'd need a LLM back tracking constraint solver and I think the bottleneck will be good grading rules on whether a direction is actually architecturally clean or extensible. I'm not sure how
107.
by 2001zhaozhao 4mo ago
Inb4 the economy is just a paperclip maximizer, a hedonium maximizer, and 5 different AGIs built to maximally enrich their creators all trading with each other.
108.
by 2001zhaozhao 4mo ago
It's probably possible to earn a $1B as a good startup. I think this article is right that it is a lot more possible to ethically get there as a startup than through other means, which are often much more political and thus overwhelm
109.
by 2001zhaozhao 4mo ago
Well crap, I was planning on doing this. Although I did plan for OVH-level dedicated server prices, so as long as they don't jack up prices too I'll be fine...
110.
by 2001zhaozhao 4mo ago
Good writeup. I think the main difference in my workflow is that I skipped the sandboxing part and accepted the coding agent having access to the entire 24/7 dev machine, so I'm still running on worktrees. Also, the "idea enr
111.
by 2001zhaozhao 4mo ago
I think programs like this are cool, the company gets to promote their product and do good at the same time. This looks like a broader program than past ones and giving out GPT5.5 could be meaningful in improving open-source projects'
112.
by 2001zhaozhao 4mo ago
Wow, what an awesome GitHub README. Informative and straight to the point. I'm building an IDE for coding agents with a project management kanban board & plugin support as two of its flagship features. I think I can learn a lot fro
113.
by 2001zhaozhao 4mo ago
> more stuff done More stuff done per dollar or more stuff done for more dollars? Seems to be an important distinction
114.
by 2001zhaozhao 4mo ago
Thousands of Anthropic employees believing they just finished putting out fires related to Fable this week and finally won't be on call for this weekend:
115.
by 2001zhaozhao 4mo ago
GPT-OSS flashbacks intensify
116.
by 2001zhaozhao 4mo ago
I wonder whether a layering like "Human -> LLM -> Human -> LLM -> Human" would work and provide the best of both worlds if designed and enforced well.
117.
by 2001zhaozhao 4mo ago
> scaling dynamics seem to be evolving like any other org – just at a higher employee count and revenue numbers. I really wonder if it's possible to avoid these dynamics, even if you try really hard. If not, it seems to me that goal
118.
by 2001zhaozhao 4mo ago
individual contributor
119.
by 2001zhaozhao 4mo ago
How do they detect whether an experiment being done on a smaller model is used to improve a competing frontier model, or just an innocuous hobbyist LLM experiment?
120.
by 2001zhaozhao 4mo ago
We'll need a lot of good summarization techniques to cut down on the cost of this model. I expect that a common use of Fable 5 is to just do high level direction while delegating literally all work (exploration and implementation) to O

More ›