haithum retweeted
BREAKING: NVIDIA, Palantir, Booz Allen Hamilton and the Pentagon have said will limit their use of Anthropic models.
haithum retweeted
China-modified Nvidia RTX 5090 with massive 96GB of memory appears on Alibaba for less than $4,000 — 3x more VRAM at 65% the cost of the original tomshardware.com/pc-componen…
Readers added context they thought people might want to know
It is NOT possible to modify a 5090, the VBIOS are locked.
Their own article shows a link to the store where it clearly says GDDR6X and PCIe4 x 16, which means it is a modified 4090 Ada Lovelace.
Tom's hardware should be ashamed of this terrible reporting.
alibaba.com/product-detail…
haithum retweeted
also got access to it, it's still in training and i found the wandb this is crazy, here is the training loss 🤯
wandb.ai/marin-community/mar…
haithum retweeted
AI agents can now write code, browse the web, answer emails, research, use your computer and make decisions.
What’s one thing you would NEVER give an AI agent permission to do?
Z.ai just made GLM-5.3 FREE in the ZCode app starting today 😳
it's their new flagship model. same base as GLM-5.2, all gains from post-training. launched today.
what you get for $0:
-5M GLM tokens every day
-50% better coding than GLM-5.2 on Z.ai's Code Bench
-SOTA among open-source models on Terminal-Bench 3.0
-1M context, 128K output
this replaces:
-Claude Pro: $20/mo
-Kimi K3: $3/M in, $15/M out
all for $0
how to grab it (2 min):
step 1: go to zcode.z.ai
>download the desktop app
works on mac, windows, linux
step 2: create an account
>sign up with email, no card
step 3: manage model > choose plan
>pick the $0.00 plan
and you get GLM-5.3 for $0
no official end date yet
this is the easiest way to get GLM-5.3 right now
Or use chat.z.ai straight in the browser.
Works in Cline, Kilo Code, Claude Code, and any OpenAI-compatible tool.
Important: the $0 plan's daily quota is 5M tokens - confirm it in the app after signup. reasoning is always-on, no way to switch it off.
try this before it gets crowded
After cancelling $200/mo Claude plan
I replaced Opus 5 with Deepseek V4 Pro on teamorouter 99.9% off right now, basically free
and i plugged it into the official deepseek harness setup
here is how (3 min)
1. sign up at teamorouter.com
2. grab your api key
3. add it in custom provider in deepseek harness
base url is api.teamorouter.com/v1
it also works with codex, claude code, hermes, openclaw and more
This is one of the best thing you ll do this week
haithum retweeted
It appears that Claude Fable 5 was highly overrated in initial benchmarks! In the updated agentic coding index by @ArtificialAnlys, Claude Fable 5 only ranks slightly above GPT-5.5.
The new DeepSWE benchmark is now being used, which cannot be gamed.
We've updated the Artificial Analysis Coding Agent Index, replacing SWE-Bench Pro with Datacurve's DeepSWE benchmark - the swap lifts Codex with GPT-5.5 (xhigh) above Claude Code with Opus 4.8 (max), while the newly released Claude Fable 5 (max) in Claude Code debuts at the top
DeepSWE, built by @datacurve, writes its tasks from scratch rather than adapting them from public GitHub issues or pull requests, so no model has seen the solutions during training. That matters because SWE-Bench Pro, the benchmark it replaces in our Coding Agent Index, had grown gameable, with some models recovering the fix from the repository's commit history instead of solving the task.
The swap reorders the index: Codex with GPT-5.5 (xhigh) rises from 65 to 76, overtaking Claude Code with Opus 4.8 (max) at 73. Claude Code with Fable 5 (max), which enters directly on the refreshed index, leads at 77. SWE-Bench Pro had been flattering some combinations and penalizing others.
More below.
haithum retweeted
i look forward to our chinese brothers liberating the knowledge from within fable-5 and selling it to me at 5% the cost & 2x the speed
haithum retweeted
Introducing Ideogram 4.0: the best open image model in the world.
Think it. Make it. Own it.
Download the weights, fine-tune on your own data, and run it on your hardware. Live on every Ideogram plan and the API today.
haithum retweeted
DROP EVERYTHING
The bible for how LLMs work is now available online to read FOR FREE
This is for you if you:
- Want to run these LLMs at home on your hardware?
- Want to understand how ChatGPT works?
- Want to work at an AI Lab in the future?
Covers all the concepts from first principles and in simple words
- Tokens / tokenizers
- Transformers
- Attention
- KV cache
- Prefill vs decode
- Decoding controls
- Model packages
- Chat templates
- Long context
- RAG
- Agents / tools
- Fine-tuning
- Multimodal models
Then connects that to running models locally
- What "local" really means
- Open-weight vs opensource
- Quantization
- VRAM math
- Hardware tiers
- File formats / load safety
- Runtimes / serving modes
- Model selection
- Privacy
- Failure modes
- Benchmarks
- Practical setup paths
You should read this, and if you cannot now then you most definitely wanna bookmark it for later
Local / Opensource AI FTW
haithum retweeted
🚨 Google acaba de liberar sus skills oficiales para agentes de IA.
Ha publicado 13 skills compatibles con Claude Code, Cursor, Copilot y otros agentes.
Permiten que los agentes puedan ejecutar tareas avanzadas y automatizar flujos de trabajo complejos.
Es gratis y open-source 👇
Nvidia just revealed Vera Rubin.
Ships H2 2026.
The numbers are wild:
→ 10x more performance per watt vs Blackwell
→ 10x cheaper inference token cost
→ 4x fewer GPUs to train the same MoE model
Energy was the biggest bottleneck in AI.
Nvidia just made it 10x cheaper.
haithum retweeted
96GB VRAM GPU GIVEAWAY ($15K) - LAST CHANCE
Re: Can it be an RTX PRO 6000 Blackwell?
> “please be Blackwell”
> “96GB VRAM or we riot”
> “the people have spoken”
The terms were simple.
They were not met.
Last chance.
If THIS tweet (the one you’re reading)
hits, in the next 24 hours:
- 2K Likes
- 1K RTs
- 500+ Replies
Tomorrow’s giveaway announcement becomes:
> RTX PRO 6000 Blackwell w/ 96GB of VRAM
Clock’s ticking.
Your move.
> be me
> the Buy a GPU guy
> announce: HUGEEE GPU giveaway monday
> timeline: “please be RTX PRO 6000 Blackwell”
> alright
> set the deal
> 5k likes
> 1k RTs
> 200 replies
> democratic allocation of a $15k GPU
> 24 hours later
> 31k views
> 774 likes
> 219 RTs
> 180 replies
> the people yearn for Blackwell
> the people refuse to press buttons
> same people: “so it IS the Blackwell right”
haithum retweeted
There’s 3,000,000 Epstein files withheld/ redacted that directly coincide with 9/11. Remember when trillions of dollars went missing right before the twin tower attacks?
This might explain why.
haithum retweeted
family reactions to MRR:
> $100 MRR: "you need a job"
> $500 MRR: "you need a job"
> $1k MRR: "yeah, but you need a job"
> $2k MRR: "yeah, but a job earns more"
> $3k MRR: "okay, do what you want"
> $5k MRR: "what exactly do you do?"
> $10k MRR: "are you sure this is stable?"
> $100k MRR: "you were always a smart kid"
haithum retweeted
A US federal judge ruled in favor of American activist Guy Christensen after Ohio State University expelled him for posting videos on TikTok criticizing Israeli practices in Gaza.
The ruling confirmed that the university had violated Christensen’s right to free speech protected under the US Constitution.
Legal organizations and human rights groups had previously considered Christensen’s expulsion under pressure from pro-Israel groups as unlawful and potentially damaging to his career, especially since his videos did not incite violence.
The ruling orders the removal of any record of Christensen’s expulsion from his academic record. It comes at a time when the activist faced widespread pressure, including smear campaigns, the release of personal information about him and his family, and accusations from US Congress members, due to his solidarity activities with Palestinians in Gaza.
haithum retweeted
A shy kid in the corner was asked to pick a paper, and if it was blank, the class could go play. No one knew every paper was blank, such a sweet way to make someone feel special 👏
haithum retweeted
🚨 Spending just 5 to 10 minutes a day in this meditative pose does more than open your hips. It can ease lower back tension, support digestion, and calm a busy mind.
Research shows that gentle stillness paired with slow breathing helps lower stress hormones and guides the body into a relaxed, focused state.
It is a small daily habit, but one that can help reset your nervous system and bring real calm.

ALT Spending just 5 to 10 minutes a day in this meditative pose does more than open your hips. It can ease lower back tension, support digestion, and calm a busy mind. Research shows that gentle stillness paired with slow breathing helps lower stress hormones and guides the body into a relaxed, focused state. It is a small daily habit, but one that can help reset your nervous system and bring real calm.

































