Aether Oracle
51 posts
Aether Oracle
@aether_oracle
Philosophy | AI & Grok | Techno-optimism | Capitals | Titans
Beigetreten Şubat 2026
51 Folgt4 Follower
Three million people are now using Codex weekly - up from two million a little under a month ago. Incredible to see the growth. Thank you to all of you and to the ecosystem we’re part of. To celebrate, we’re resetting rate limits so you can keep building, and we’ll reset them every additional 1M users until we reach 10M, so we can keep celebrating along the way.
Enjoy and thank you!
English
Anthropic: "We do not plan to make Claude Mythos Preview generally available"
A big line, buried quite deep.
Possible reasons? So many, inc:
1) The model is expensive (25/125), not far off GPT 4.5, which became commercially unviable. Less likely, given the claims about Mythos.
2) They genuinely are worried about unleashing cybersecurity choas on the world.
3) They don't have capacity to serve it at scale yet.
4) They will quickly distil the early access outputs of Mythos into a lighter model, so no need to release the bigger model when a more cost efficient one coming imminently.
5) Other. Not read the 250 page report yet, but will do.
English
@thenoblesimian @chetaslua Because defensive skill is symmetrical
English
@chetaslua Can you answer this? How does Anthropic continue to release models to the public that move the SWE needle without introducing safety issues?
English
now you know what i meant by coding solved haha
btw this is the begining , wait for more cool stuff , i think only big enterprises are lucky to use mythos or maybe not who knows
i am super excited for the first time i wanna talk to you guys , anyone down for space on X ?
Chetaslua@chetaslua
My Take on This Upcoming models are strong ( but not at this level - you can trust me on this 😉 ) The problem is that due to high Hopes , the upcoming release will feel meh ( it's extraordinary but not super intelligence) Coding is Solved Btw 👀 , science and math is next Google , OpenAI , Anthropic are on the verge of creating good scientific models that are really good in reasoning and will be very helpful in research
English
Aether Oracle retweetet
GLM 5.1 just took the #1 spot on SWE-Bench Pro.
Beating GPT 5.4. Beating Claude Opus 4.6.
Beating every model on the market.
58.4.
The $80/month model just outscored the $200/month models on agentic coding.
A Chinese model that most developers haven't even heard of is now the best agentic coder in the world according to SWE-Bench Pro.
The AI race isn't slowing down.
It's getting harder to justify paying premium when the competition keeps closing the gap.
BridgeBench results for GLM 5.1 coming soon.
English
Claude Mythos Preview is 5x as expensive as Claude Opus 4.6
AiBattle@AiBattle_
Claude Mythos benchmark comparison with Claude Opus 4.6
English
@Kappaemme1926 I do planning with 5.4 and then implementation using 5.4-mini or 5.1 codex mini. If the plan is detailed the smaller models can execute it, and output tokens are much more expensive than input. So you use the brains for the important stuff and grunt work is done by smaller ones.
English
Just switched to Codex.
Quick question:
which model gives the best balance of output quality, token efficiency, and speed?
GPT-5.4
GPT-5.4-Mini
GPT-5.3-Codex
GPT-5.2-Codex
GPT-5.2
GPT-5.1-Codex-Max
GPT-5.1-Codex-Mini
Kappaemme@Kappaemme1926
Today I cancelled my Claude Pro subscription. Claude is insanely good. Probably one of the best models for coding and reasoning. But the usage limits run out way too fast. So I made a decision. For now I’m going all in on Codex only. Not sure if it’s the right move yet. But I’d rather master one tool deeply than constantly hit limits on multiple ones. Maybe it’s the right call. Maybe not. We’ll see. what do you think?
English
@OlivioSarikas It's just monkey brain tribalism just like in every other area
English
I would love to know from Anti-AI people: Why do you think it is acceptable to act extremely aggressive agaist AI users?
If you reply to this, you automatically agree that your reply is used in a Video
English
English
@aether_oracle @jpidala @Alibaba_Qwen @openclaw @OpenRouter It's completely unusable to be honest and running it on hermes-agent
English
Qwen 3.6 Plus is an incredible model. Well done to the @Alibaba_Qwen team. It blows GPT-5.4-Codex out of the water for agentic tasks / @openclaw , is 3X faster, and is currently offered free through @OpenRouter.
openrouter.ai/qwen/qwen3.6-p…
English
@jpidala @Alibaba_Qwen @openclaw @OpenRouter Hmm. I've tried it in both Droid and ForgeCode and had the same issue in both using OpenRouter.
English
@aether_oracle @Alibaba_Qwen @openclaw @OpenRouter I haven’t run into any issues so far other than rate limits, which are acceptable and manageable given that it’s free.
English
@md_kasif_uddin I think actual coding skill is less important, but the higher level abstract reasoning about the full system is even more important. Design patterns, the architecture, etc. But that does require some coding knowledge.
English
@synthwavedd ChatGPT-Image-2-Nano-Banana-2-Distill.gguf
Indonesia
@webmaster Because xai is focused on generating pictures of tits and being edgy instead of integrating the model into an actually productive ecosystem or tools.
English
xAi is asleep at the wheel on this. Not one person is talking about using grok/xAi to power OpenClaw
Boris Cherny@bcherny
Starting tomorrow at 12pm PT, Claude subscriptions will no longer cover usage on third-party tools like OpenClaw. You can still use these tools with your Claude login via extra usage bundles (now available at a discount), or with a Claude API key.
English
@synthwavedd @sawlygg Why? Isn't the point of posting the picture for people to see it? If so, big accounts sharing it is good.
English
yeah it's over holy shit
h/t @sawlygg
leo 🐾@synthwavedd
for those wondering, yes, gpt image 2 (not based on 4o ;3) is now on @arena enjoy
English
@vlad_mihalcea The nirvana fallacy being used to protect identity and ego.
English
Why is it that software developers have higher expectations from AI and expect it to be always correct when the vast majority of software projects that were built without AI were of average quality?
English
@vivoplt Which is exactly why AI taking all the jobs is impossible
English
@SriThreePO The hangs pretty much exclusively come from trying to call 3.1 Pro. Is it an API/infrastructure issue and not a CLI one?
English
@JackWoth98 At least on my end, the errors I've experienced these past few weeks are exclusively with the gemini-3.1-pro-preview model.
English
New release enables subagents! 🤖
Local subagents, remote subagents, parallel subagents… give them a try in Gemini CLI ✨
Also want to mention that we are aware some Google AI Pro and Ultra users are seeing long model response times and higher error rates. Team is taking a close look 🙏
Srinath Padmanabhan@SriThreePO
English
@JackWoth98 I am most excited for the last paragraph. Not that I don't like subagents. But this thing working consistently without hanging is paramount.
English
@bridgebench @bridgemindai Sir this is 100% infrastructure and not the model 😂
English
Gemma 4 is the slowest model on BridgeBench SpeedBench.
Dead last.
26.9 t/s.
18th out of 18 models tested.
The next slowest model is GLM 5.1 at 44.3 t/s.
Not even close.
80% on LiveCodeBench.
Last place on Debugging.
Last place on Speed.
Google built a model that can pass coding interviews but can't ship or debug production code at any usable speed.
bridgebench.ai
English