Fabio Builds

@MeinAjeebHun @Fabiobuilds @Bhavani_00007 the bottleneck isnt electricity its the cost of GPUs required to run this stuff in a decade you can probably get an H100 performance for your home and run your own LLMs, so its basically free or 10$/mo

@HeyAbhishek If this is true it will be a big game changer given Claude Design limits to around 3 hours of work per week on 200$ plan. But I’m afraid with real usage we are still far from Anthropic. Worth a shot.

Open Source is actually winning this time! Kimi just dropped K2.6, China's answer to Claude Opus4.7 it totally cooked claude design🤯 and it's 7x cheaper and 100% open source see prompt in the comments👇

@LexnLin Every time o tested Mistral it turned out as a huge waste of time. Hope to change my mind soon

NEW AI MODEL FROM EUROPE 😭 no comeback imo but open weights is nice

@bridgemindai The only improvement we were waiting for, let’s wait for a confirmation and ideally a note from them

Claude Code rate limits are way better now. I'm at 27% of my session limit and I've been vibe coding all morning with Claude Opus 4.7. 23% of my weekly limit used. A week ago I was hitting limits in 2 hours. Anthropic actually fixed the rate limits.

@spicey_lemonade Must be benchmaxed, it’s actually unusable for coding in real business

Gemini 3.1 is in the top 3 of almost every main benchmark, yet no one uses it. I think vibecodebench, swe atlas, and AA agent index are well calibrated.

@Bhavani_00007 With the price increases that Claude is doing to its models, very soon a junior dev will be cheaper than Claude code

It’s so over for junior engineers 😭 We have a new CEO at the company I’m working in, and he recently pushed everyone to adopt a full Claude-based workflow. Now everyone is using Claude for almost everything. During a meeting, he asked if anyone had questions. An intern asked whether the company would continue hiring interns or junior engineers. He said everyone currently working is safe, but they've decided to stop hiring interns and junior engineers going forward. They don’t want to spend 3–4 months training freshers anymore, and would rather hire experienced engineers who already know these tools. Are junior engineers actually getting replaced??

Was the best for really short time end of last year before the release of Opus 4.5. Then it seems they just quit the run and started focusing on small and local models, where they are actually doing great job (flash/ lite models are surprising me more and more, but cannot code properly)

genuine question not a single person i have ever spoken to uses gemini for coding. this is still very very weird. why is gemini so bad at coding when google has scoured the web full of code for decades?

@Star_Knight12 It’s over for any hard skills that it currently non relying on agentic coding. It’s the same that happened with computer half century ago, they just need to change their toolbox

is it over for 3D artists

@vivoplt Switching to Codex

Is there any way to use Claude Code without hitting the usage limit

@om_patel5 who uses gitHub Copilot when you can get the same outcome basically in a native way with codex / Claude? Considering price is converging to the same, I do not see any valid use case for it

GITHUB COPILOT JUST RAISED AI MODEL PRICES BY 9X OVERNIGHT starting june 1 2026, every annual copilot pro and pro+ subscriber gets hit with the new multipliers across both claude AND codex models for claude: > claude opus 4.5 went from 3x to 15x. > claude opus 4.6 went from 3x to 27x. > claude opus 4.7 went from 7.5x to 27x. > claude sonnet 4.5 went from 1x to 6x. > claude sonnet 4.6 went from 1x to 9x. > claude haiku 4.5 stayed at 0.33x (the only one untouched) for codex/openai: > GPT-5.1 went from 1x to 3x. > GPT-5.1-codex went from 1x to 3x. > GPT-5.1-codex-max went from 1x to 3x. > GPT-5.2 went from 1x to 3x. > GPT-5.3-codex went from 1x to 6x. > GPT-5.4 went from 1x to 6x. > GPT-5.4 mini went from 0.33x to 6x (18x increase) > gemini 3 pro went from 1x to 6x. > gemini 3.1 pro went from 1x to 6x. if you locked in an annual plan expecting stable pricing, you just got the rug pulled and if you were running opus 4.6 or codex max inside copilot at scale, your monthly rate just multiplied by up to 9x this is the third quiet pricing change across the AI industry in two weeks: > anthropic locked opus behind a paywall-within-a-paywall for pro users > anthropic removed claude code from the pro plan > now this from github every AI subscription is FORCING us toward usage billing whether developers like it or not your $20/month all you can eat plans are getting ripped apart with usage caps and higher billing the era of cheap AI tooling is officially ending it was nice while it lasted though

@DavidOndrej1 As if it wasn't already bad enough 😅

they have nerfed Perplexity Computer

@BannedLatino @Hesamation let's see. I would be very happy if this turns out to be true and I can avoid the continuous price increases of US models; but so far, my experience with DS and other Chinese models have been quite unreliable

@Fabiobuilds @Hesamation Dude, was released day ago and you writing this BS, I was coding with v3, now v4, is fixing all the bugs very fast, you are not even using this. Lmao

you don’t realize how CHEAP DeepSeek is until you use it all day and pay the price of a bag of chips. DeepSeek V4 Pro: ~$0.87 / 1M tokens Opus 4.7: $25 / 1M tokens that's ~28x cheaper. Opus is stronger on benchmarks, but at this price gap, stronger has to be A LOT stronger.

@flynnrobinson_ @ClaudeDevs I think it’s a feature rather than a bug, I always feel good when I ship in 1/10th of the estimate time

@ClaudeDevs Now fix the system peompt that's encouraging claude to say it will take 3 weeks to do something, and then does it in 2 minutes. Also, the number of times it's telling me to go to bed or take a break

In the last four Claude Code CLI releases, we’ve shipped 50+ stability and performance fixes. Faster resume, stable auth, lower memory, fewer hangs: 🧵

did Anthropic launch Claude Music or something

@jamesrichardfry Till they soften the limits of Claude design (or a real competitor appears). Then all engineers will be designers

in the era of ai all designers can ship good code few engineers can ship good design

Rate limits have become so strict to make it look quite unusable at most plans. At the same time it got nerved, so you would need it to solve its own mistake and get rate limited in the middle of the debug. At the same time, codex become a strong quality contender for a fraction of Claude price

genuine question if Codex can write features review code and even open PRs for you what exactly is your role as a developer ? are you still building or just supervising ?

@vlelyavin @bridgemindai Wow, true. I can also just imagine the cost of letting the model run for 2 hours

@vlelyavin @bridgemindai Does anyone even use gpt pro? I mean, it’s like 18 times more expensive than Opus 4.7, trying to understand what are real use cases

GPT 5.5 is now the most expensive frontier model in the world. $11.30 per million tokens. Claude Opus 4.7: $10. DeepSeek V4 Pro: $2.20. Kimi K2.6: $1.70. A few months ago Sam said Anthropic was the expensive product for rich people. Today Anthropic is 13% cheaper than OpenAI's flagship. The expensive product is yours now, Sam.

@Hesamation The real cost of deepseek is the time you spend debugging its mistakes before going back to codex / Claude

@xoaanya Claude starts making sense at 100$, it’s unusable below that. At 200$ you can start thinking of real work. Codex lets you work already at 20$, but quality is up for debate

Now that GPT-5.5 is out, would you still pay $20/month for Claude or Codex?