Sample (@sample) header
Sample (@sample) profile picture

Basketball, American democracy, and Computer science

Joined February 2007
Photo by Sample (@sample)Photo by Sample (@sample)Photo by Sample (@sample)Photo by Sample (@sample)Photo by Sample (@sample)Photo by Sample (@sample)Photo by Sample (@sample)Photo by Sample (@sample)
Musk makes verification a false flag devoid of meaning and then spends verification resources hunting down false authenticity? What kind of nonsense... ? 🤔🧠💡
You’re unable to view this Post because this account owner limits who can view their Posts. Learn more
2
2
6
Google has been a Cluster-F since Argon. Naming, subscription value, speed, releases... things they were good at! Pro and Flash were great. The quotas were generous. Since Argon the only people hyping are shills or dunces, and Google's release communication is atrocious!
tf is Gemini 4 Neon Never thought I would say it but there are too many models
1
7
Gemini models look like they're desperately fighting the instinct to turn any design request into a modification of a dashboard full of micro-text and live icons in every pixel. I'm convinced it trained/wants to make human anatomy diagrams and dashboards and not webpages or apps
Replying to @HarshithLucky3
You mean frontend development? Have specific feedback on it? Aesthetics or functionality?
11
Gemini models look like they're desperately fighting the instinct to turn any design request into a modification of a dashboard full of micro-text and live icons in every pixel. I'm convinced it trained/wants to make human anatomy diagrams and dashboards and not webpages or apps
7
Replying to @melvinjohnsonp
The biggest problem: Anthropic's B2B (API) subsidizes the subscription, providing value for it, positive sentiment, and market share. Google's B2B (TPUs) places the lack of compute, onto bad quotas and bad subscription value for users, hurting value of service and poor sentiment
1
This chart was tested and released this week. I'm not sure if you're making things up with "nuance" but please site your sources if you think you know of better evidence for your claims. This result looks thorough, and each test was run multiple times to confirm results.
6
I'm not sure if that's all accurate. Also isn't the API price of Opus twice as much? Anyways here is a recent, posted today, empirical evidence test. His conclusion: "The obvious winner is Opus High. But Sol is close." teapawt.org/PovilasKorop/status/21…
New chart on my AI Coding Daily website. LLM score from leaderboard with subscription usage % after 20 prompts. The obvious winner is Opus High. But Sol is close. In fact, on Medium level, they are almost identical in score/usage. Full page: aicodingdaily.com/compare/ef…
1
1
13
Replying to @PovilasKorop
Wow, this puts into perspective how bad Google Gemini's offering is. Everyone knows it, but I'm shocked that Hydrogen is even so bad value to points ratio. Imagine using 6% of your subscription for a 73 quality instead of 10 points higher for the same value. 🤦‍♂️🤦‍♂️ Imagine Argon
12
Replying to @loftwah
Yep. But Sol 6.1 is super comparable with better quota, so it's just a choice. The nice thing is too that they're the same price, so its easy to go choose and use Opus 5.5
1
1
19
If you think it was a bug in inference or something you could file it with them with the chat id
1
1
8
Replying to @loftwah
It's because Sol 6.1 is one of the best models out, and usable at amazing subscription value. Two things Google hasn't even accomplished yet. They're not behind when they were that far ahead
1
1
19
Claude uses their enterprise API users to subsidize their subscription into fantastic marketshare, sentiment, and value. Google Gemini team uses their enterprise users (TPUs) to have no compute and make their subscribers pay for that lack, in atrociously bad subscription value.
21
Replying to @kunchenguid
you're right, but you're also overlooking that their subscription is the most hostile even simply because of the aggressiveness of their usage limits. it's by far the worst *value* subscription to subscribers, and Pro users get like 6? tasks of the Pro model (Argon) per week. ‼️
18
Replying to @CoachHackGO
I had a youth team here in Europe running circle into flex because it fit so naturally and the coach wanted continuing options off it. We also had a rule that the butt-cutter, if switched and stopped/blocked immediately, triggered the "butter" (screener) to cut instead.
1
1
14
Replying to @LuminaBench
Google's rather atrocious post-Argon release, communication and confusion, and subscription plan with Pro getting 6 pro queries exhausting weekly subscription, still puts them squarely in the "Google of AI" category.
12
In fact, what's crazy, is it's that Google's enterprise business selling TPU's is exactly what has hurt their Gemini subscription marketshare and interest, and made it the worst value subscription money can buy, by far.
1
11
Replying to @matt_teeixeira
To be honest it's a great strategy. Google is at the far opposite end offering an atrocious subscription and the sentiment on Twitter has highly negative towards their offering and comms since Argon. Argon still offers like 6 prompts exhausting Pro's weekly usage, which is insane
1
1
16
Maybe that's exactly it. The enterprise API payers are subsidizing the subscription for market gain. 🤔
How can Claude afford subsidizing the $200 plan? We’re getting 60x the API price.
1
1
40
Replying to @PawelHuryn
Oh wow, I had no idea Terra was that bad
11
Replying to @Lentils80
In general web development does it still look like it's struggling to not make everything a variation of an incredibly over-cluttered dashboard full of live-indicators and micro-text everywhere?
86
Doesn't Google have objectively the worst subscription? Yes, you get more when you pay more, but the value stays poor. Haven't all empirical tests and benchmarks (I can link you if you want) shown it to be by far the worst way to spend money in both value and performance?
1
1
50