Christos Nikitopoulos retweetledi
Sair do cinema de noite, depois de um filme que você gostou muito é uma sensação do krlh mesmo
floptok@_eoantao
Como é quando você sai do cinema e tudo parece irreal
Português
Christos Nikitopoulos
123 posts
@xai I am sure there will be a release for pro members.. I would try for three days to test out quotas and such if its free for three days, but don't think $300 is something affordable for me at this time.. Maybe a one time purchase to test
English
I hope Gemini models can keep up the performance.I have noticed performance drops month after month with Gemini in Antigravity, and in some cases the app. However, I do genuinely believe at one point or another there will be a medium of substantially good results.
Bindu Reddy@bindureddy
Gemini 3.2 Flash - Capitalizing on DeepMind's clever distillation techniques... Rumors are that benchmarks show it's hitting 92% of GPT 5.5's performance on coding and reasoning tasks while being 15-20x cheaper on inference costs. The latency improvements are insane - sub-200ms for most queries. Google's distillation + sparsity techniques are paying off massively. They've essentially compressed a frontier model into a flash variant without the usual quality cliff.
English
@weswinder Its kind of odd, because in the mean time they also announced 50% more weekly limits from now until July 15. At one point or another, they will need to compromise.
English
anthropic just downgraded claude code subs
programmatic usage previously used your full rate limits
which was something like a 5-10X discount vs api costs
now ALL programmatic usage is capped at $20/$100/$200 in api costs (a single opus session can easily burn $20)
not good
ClaudeDevs@ClaudeDevs
Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage. The credit covers usage of: - Claude Agent SDK - claude -p - Claude Code GitHub Actions - Third-party apps built on the Agent SDK
English
Christos Nikitopoulos retweetledi
i think a lot of people quietly give up on their life because they become “realistic” too early. like they stop themselves before life even gets the chance to test them. but some people stay so delusionally optimistic about their future that they refuse to let their current reality become their final story. that’s probably why Elon Musk kept going even while rockets were failing publicly and Tesla looked finished. maybe that mindset changes everything. the ability to not emotionally collapse every time life refuses to validate you. that level of optimism changes people. because when you deeply believe something is possible, you behave differently. you last longer. you recover faster. you keep going during the phase where most people quietly lose belief in themselves.
alexei@alexeixbt
normalize realizing that the whole cheat code to life is being insanely delusional and optimistic
English
@mark_k What would be your recommended solution for this?
English
@mark_k I was thinking the same thing. Going from a mathematical and scientific context mindset, to more of a "creative writer" has been in a weird way problematic.
English
It’s supremely ironic that good writing turns out to be one of the hardest problems for "Large Language Models".
English
Current State of AI for Coding:
Claude: Expensive and frequently hitting quotas.
Gemini: Experiencing "Antigravity" server errors, with integrated models underperforming. (Positive reviews on AI Studio.)
Codex: Positive reviews for now, though quotas are expected to drop by a factor of 2 by the end of the month.
Grok: Rolling out new build features and enhanced computer usage capabilities.
Who will emerge as the winner? The one that delivers the most quotas per dollar while maintaining strong token efficiency, based on subscription value. I think if Grok can provide significantly more quotas, it could set a new standard that competes effectively against the other companies. Time will tell for sure, but right now it seems pricing feels overwhelming for the majority of consumers
Genuinely, what are your thoughts?
English
My biggest wish for @xai (SpaceXAI) right now is simple:
Please build a powerful new image model that can genuinely compete with Nano Banana 2 and GPT-Image-2. Grok Imagine is already fun and creative, but xAI needs a real flagship image model next. 🙏
English
@elonmusk Hi Elon,
Hoping this reaches you for the Grok community—it would mean a lot. A reply would be amazing, but even just reading it counts.
I’ve seen almost no YouTube videos or broader buzz about the new Grok CLI, build, and computer features—mostly just discussion on X. I get that they’re fresh and likely being tested by the heavy users.
Any sense of when pro members might get access in the near future?
Thanks,
Christos
English
@blankspeaker Based on your intake so far, it seems that Grok build is based on subscription rather than API? Is this an accurate assesment?
English
Grok Computer is rolling out now .
Basically it is a storage location where Grok will save the files it creates .
Blob Fish@Blob_Fish_19
Grok computer is here?!
English
This is genuinely another concern of mine for the economy as a whole especially with the gigantrium amount of money that is flown to investing into AI.
If the chinese models are 6-12 months behind, and a majority of people are able to have open source models as powerful as mythos installed locally, then what would happen to subscription based models if people know they can run it locally?
The only thing that would prevent a mishap of investments towards subscription based models after, is that the demand of local infrastructure doesn't run nearly as high as that of cloud models. For instance, this kind of reminds me of the Steve Jobs, and Wozniak.
Wozniak had the thought that people would love to build custom pcs, while Steve Jobs on the other hand had the notion of selling pre-built. We know who won the majority of the consumer market (Jobs).
In a nutshell, I answered my own impulsive thought of initial fear. 😅
vitrupo@vitrupo
Dario Amodei says Anthropic has to release Mythos widely enough to harden cyber defenses, but carefully enough not to spread the threat first. Too few actors, and defenders stay weak. Too many, and misuse scales. Too slow, and Chinese models catch up. Too fast, and the safeguards may not hold.
English
I think its also possible, that once Grok receives demand with their high quotas with Grok build, that if subscriptions or APIs surpass the revenue from the lease then:
Cut off lease, meet demands + demand now being higher thus already making more profit then the lease. Competition now is weaker due to far lesser compute.
Gergely Orosz@GergelyOrosz
So let me get this right: 1. Anthropic bans xAI from using Claude (to stop them from perhaps distilling Claude for their own model) (...) 2. xAI gives up ~a quarter of its DC capacity for Anthropic to rent and run Claude A win for Anthropic no doubt. What's in it for xAI tho?
English
Imagine Grok 5 comes out and surpasses all AI companies, even after giving Anthropic Colossus 1.
English
@ChrisWitme87550 @MarioNawfal @elonmusk @grok sick him, put him in his place ... unhinged 🤣 (testing to see if this actually works)
English
@MarioNawfal @elonmusk EV’s are more harmful to the environment lol. You people really do think we are retarded.
English
One of Elon's most vocal critics just bought a Cybertruck.
Brian Krassenstein, who has spent years publicly clashing with @elonmusk, announced the purchase yesterday.
His reason had nothing to do with politics.
He has a young family and the Cybertruck is the only pickup truck in America to hold both an IIHS Top Safety Pick+ award and a perfect 5-star NHTSA rating simultaneously.
When your fiercest critics are buying your product because the data leaves them no choice, that's a different kind of win.
English
@teslaownersSV @grok what does if bench exactly test?
English
BREAKING: xAI's Grok 4.3 takes the #1 spot on IFBench with an 81% score, according to Artificial Analysis.
It beats every major competitor including GPT-5.5, Gemini 3.1 Pro and Claude Opus 4.7.
Muskonomy@muskonomy
NEWS: xAI's Grok 4.3 takes the #1 spot on IFBench, the leading instruction-following benchmark, according to Artificial Analysis. Grok scored 81%, beating every major competitor: MiMo-V2.5-Pro at 80% Gemini 3 Flash at 78% Gemini 3.1 Pro Preview at 77% GPT-5.5 (Xhigh) at 76% GPT-5.4 (xhigh) at 74% Claude Opus 4.7 (max) at 59% This is Grok's latest top finish on the benchmark, which measures how closely a model follows user instructions.
English
I may be wrong, but the cost of running Grok 4.3 the cost/intelligence index is actually good, no? I know that benchmarks dont mean everything, but has to have some tangible evidence to it?
Theo - t3.gg@theo
To be fair, suing OpenAI does seem like a more viable strategy than pretending Grok 4.3 is any good
English
I do care, thanks for sharing. I definitely think some chess moves are happening here.
With the inclusion of having Grok connectors, now we have Claude coming to Colossus, partnership with Cursor.. My oh my, things are moving into a historically interesting pivot point.
Elon Musk@elonmusk
Same here. By way of background for those who care, I spent a lot of time last week with senior members of the Anthropic team to understand what they do to ensure Claude is good for humanity and was impressed. Everyone I met was highly competent and cared a great deal about doing the right thing. No one set off my evil detector. So long as they engage in critical self-examination, Claude will probably be good. After that, I was ok leasing Colossus 1 to Anthropic, as SpaceXAI had already moved training to Colossus 2.
English