Niche Publisher Nero (@LavenderScreams) header
Niche Publisher Nero (@LavenderScreams) profile picture

I wear many hats in publishing, media buying and selling, vibe coding apps, and signing up for Muay Thai courses in Thailand whenever I'm there.

A few cities every year, Goa
Joined May 2021
Photo by Niche Publisher Nero (@LavenderScreams)Photo by Niche Publisher Nero (@LavenderScreams)Photo by Niche Publisher Nero (@LavenderScreams)Photo by Niche Publisher Nero (@LavenderScreams)Photo by Niche Publisher Nero (@LavenderScreams)Photo by Niche Publisher Nero (@LavenderScreams)Photo by Niche Publisher Nero (@LavenderScreams)Photo by Niche Publisher Nero (@LavenderScreams)Photo by Niche Publisher Nero (@LavenderScreams)Photo by Niche Publisher Nero (@LavenderScreams)
One good thing still with Opus 5.5 is you can easily do to and fros with it in the middle of a long running chat, rest assured it doesn't forget its goal and is persistently working towards it. It feels responsive and you do not feel like backing out from saying something.
9
Claude weekly limits on the ultra plan seem massively nerfed completed to a month ago. Not even five hours of my normal usage and it's swallowed 13 percent of my quota.
2
32
Opus 5.5 is an amazing model. It just gets you and does what you ask. Feels super lean, doer without sounding too eager.
35
It's tuesday so got to be fable 5.5 day
213
MiMo 2.6 on OpenCode Go has glitched three times today, running the same command hundreds of times in a loop until I manually stop it. Really liking the model so far, but this is one launch-day bug that needs fixing.
1
64
1/8 DeepSeek V4 just got a lot better — and a lot more expensive. V4-Pro-0813 is now GA, with big gains on agentic coding benchmarks. From Aug. 16, DeepSeek is also changing its API pricing. Here’s what actually changed.
1
2
139
7/8 One caveat: these are DeepSeek-reported results. Its public Code Agent tests used DeepSeek’s upcoming harness in minimal mode at max effort, so results may differ in other frameworks. Worth keeping in mind before treating the chart as a universal ranking.
1
47
8/8 The bigger change for API users is probably routing. V4 Flash still looks very strong off-peak. V4 Pro makes more sense for harder agentic work. But at peak prices, using DeepSeek for everything becomes harder to justify.
41
How has DeepSeek V4 Pro fared for you—per expectations or not? A great combo could be to use Pro as panner and Flash as executor.
1
1
151
So I don't know how everyone thinks about this, but I used GLM 5.2 a while ago, and it was just okay by which I mean after using the new Deepseek Flash. What are your thoughts on this? Is GLM still relevant and sought after? Does the initial hype exist?
1
2
65
It's a sham when your ChatGPT Pro subscription expires mid chat and you decide to downgrade to Plus, and the app resumes working without even notifying you that continued usage will drain your Plus quota. So now you to wait another week to for the refill. What's happening Tibo?
1
102
Drop Deepseek v4 Pro already! My $200 ChatGPT subscription is done and I don't wanna renew cause of their crazy hallucinations!
2
85
That itch to buy that $200 Codex subscription again.
34
Get drunk and text your employer.
1
35