@vimiv.bsky.social
Damn. Mad respect for even saying this publicly. 🫡
January 12, 2026 at 12:09 AM
Damn. Mad respect for even saying this publicly. 🫡
I’ve had similar thoughts recently. @steveklabnik.com is maintaining rue which is over 100kLOC. I wonder if he’s experienced some of these problems.
January 11, 2026 at 11:46 PM
I’ve had similar thoughts recently. @steveklabnik.com is maintaining rue which is over 100kLOC. I wonder if he’s experienced some of these problems.
God. Getting second-hand embarrassment reading that whole thread. Who needs enemies (not really enemies - just works for this saying) when you have allies like that.
January 9, 2026 at 9:04 AM
God. Getting second-hand embarrassment reading that whole thread. Who needs enemies (not really enemies - just works for this saying) when you have allies like that.
Is this a public repo? I would love to take a look at something massive written entirely with an LLMs. 140kLOC is massive, about the size of nginx. How do you review all that code? How do you setup tests to make sure everything is tested and that claude isn't fudging the details.
January 7, 2026 at 11:57 PM
Is this a public repo? I would love to take a look at something massive written entirely with an LLMs. 140kLOC is massive, about the size of nginx. How do you review all that code? How do you setup tests to make sure everything is tested and that claude isn't fudging the details.
Same. Also, their usage limits don't make sense to me. It could be because I'm stupid but it could be them lying to users
bsky.app/profile/vimi...
bsky.app/profile/vimi...
That’s what I did. I used the Claude api key in the zed editor. My context window at the end of the session was 70k/200k which means 70000 tokens and my total cost ended up at $4.57. Shouldn’t 5 dollars buy me ~200k output tokens?
January 7, 2026 at 6:16 AM
Same. Also, their usage limits don't make sense to me. It could be because I'm stupid but it could be them lying to users
bsky.app/profile/vimi...
bsky.app/profile/vimi...
I’m guessing the authors meant “mid 80s” and then had some LLM vibe check the whole book for mistakes and then vibe approve all the edits. TBH, this is definitely a skills issue if Yegge had learned how to use these tools correctly the LLM wouldn’t have made this mistake. /s
January 4, 2026 at 7:23 AM
I’m guessing the authors meant “mid 80s” and then had some LLM vibe check the whole book for mistakes and then vibe approve all the edits. TBH, this is definitely a skills issue if Yegge had learned how to use these tools correctly the LLM wouldn’t have made this mistake. /s
Just wanted to say how much I appreciate your take on these tools. I’m usually skeptical of LLM usage, but your approach makes me feel good about them. Please continue sharing more details about your day-to-day successes and tribulations. <3
January 2, 2026 at 1:00 AM
Just wanted to say how much I appreciate your take on these tools. I’m usually skeptical of LLM usage, but your approach makes me feel good about them. Please continue sharing more details about your day-to-day successes and tribulations. <3
Omg that looks so cool. Will definitely be keeping an eye on it. I’ve been waiting for a tui to replace monodraw. Thanks for sharing.
December 23, 2025 at 11:41 PM
Omg that looks so cool. Will definitely be keeping an eye on it. I’ve been waiting for a tui to replace monodraw. Thanks for sharing.
I see this often at my workplace. I think it makes me feel like the other person thinks I haven't done my due diligence before asking the question. Also, if I wanted an LLM answer why would I ever ask a person. I hate when people do this and try to pawn it off as their own answer/voice.
December 17, 2025 at 4:28 AM
I see this often at my workplace. I think it makes me feel like the other person thinks I haven't done my due diligence before asking the question. Also, if I wanted an LLM answer why would I ever ask a person. I hate when people do this and try to pawn it off as their own answer/voice.
Damn they definitely don’t make it easy to understand the cost. I guess that’s by design. All the more reason to never buy credits. The 20 dollar monthly subscription probably makes more sense. The api costs feel way too high for it to provide any real value.
November 28, 2025 at 9:39 AM
Damn they definitely don’t make it easy to understand the cost. I guess that’s by design. All the more reason to never buy credits. The 20 dollar monthly subscription probably makes more sense. The api costs feel way too high for it to provide any real value.
That’s what I did. I used the Claude api key in the zed editor. My context window at the end of the session was 70k/200k which means 70000 tokens and my total cost ended up at $4.57. Shouldn’t 5 dollars buy me ~200k output tokens?
November 28, 2025 at 8:53 AM
That’s what I did. I used the Claude api key in the zed editor. My context window at the end of the session was 70k/200k which means 70000 tokens and my total cost ended up at $4.57. Shouldn’t 5 dollars buy me ~200k output tokens?
Having just tried Opus4.5 today, $25 per MTok is a lie. I'm at 70k (according to zed) with 4.57 dollars spent of the 5 dollars which means it's still around 75 per MTok. Don't believe these companies until you see it with your eyes. Free advertisement for them. What a scam.
November 28, 2025 at 7:43 AM
Having just tried Opus4.5 today, $25 per MTok is a lie. I'm at 70k (according to zed) with 4.57 dollars spent of the 5 dollars which means it's still around 75 per MTok. Don't believe these companies until you see it with your eyes. Free advertisement for them. What a scam.
I agree with you after seeing posts on hackernews and bsky. However, if I exclude the opinions here and on hackernews, it feels overhyped.
November 26, 2025 at 5:53 AM
I agree with you after seeing posts on hackernews and bsky. However, if I exclude the opinions here and on hackernews, it feels overhyped.
It’s okay the next model will definitely fix all the problems. It’s only a matter of time before you can click one button for the AI to come to your doorstep and grab all the tax forms. /s
November 25, 2025 at 10:11 PM
It’s okay the next model will definitely fix all the problems. It’s only a matter of time before you can click one button for the AI to come to your doorstep and grab all the tax forms. /s
How about 20 years just to be really sure that AI gets it right? Then in 30 years, we might let the quantum AI's create a quantum OS?
November 10, 2025 at 2:23 AM
How about 20 years just to be really sure that AI gets it right? Then in 30 years, we might let the quantum AI's create a quantum OS?
Is this similar to Gemini cli extensions which were added back in August?
October 24, 2025 at 10:47 PM
Is this similar to Gemini cli extensions which were added back in August?
Hey what roles are you interested in? My company is hiring feel free to dm me.
October 7, 2025 at 9:05 PM
Hey what roles are you interested in? My company is hiring feel free to dm me.
Looking forward to this. I would love to hear more about managing TUF repo and how tufaceous plays a role in the update process.
September 22, 2025 at 6:43 PM
Looking forward to this. I would love to hear more about managing TUF repo and how tufaceous plays a role in the update process.
What’s it graded at? Looks like v5-v6.
September 6, 2025 at 9:18 PM
What’s it graded at? Looks like v5-v6.
Do you ever use the template snippet feature that monodraw offers? I tried searching online but seems like nobody uses that feature.
August 25, 2025 at 9:01 PM
Do you ever use the template snippet feature that monodraw offers? I tried searching online but seems like nobody uses that feature.
Does this mean we don't trust educators and medical professionals to distinguish between good and bad LLM output?
August 17, 2025 at 1:23 AM
Does this mean we don't trust educators and medical professionals to distinguish between good and bad LLM output?
It's weird when I think about an software engineer using this tool I think oh they'll be able to distinguish when the LLM is just plain wrong (even though I've seen the opposite in practice). I guess this is why you said the medical and education should be scrutinized more.
August 17, 2025 at 1:21 AM
It's weird when I think about an software engineer using this tool I think oh they'll be able to distinguish when the LLM is just plain wrong (even though I've seen the opposite in practice). I guess this is why you said the medical and education should be scrutinized more.
If they are blatantly using chatgpt during our session, it would make me leave immediately. What if it's less blatant, like seeing the app open in another window, while they are showing me the results from a previous test. The harder it is to confirm the more it eats away at you.
August 17, 2025 at 1:18 AM
If they are blatantly using chatgpt during our session, it would make me leave immediately. What if it's less blatant, like seeing the app open in another window, while they are showing me the results from a previous test. The harder it is to confirm the more it eats away at you.