I've never been lonely. I like myself. I'm the best form of entertainment I have ,me and the doggo.
République du Bénin
Joined December 2018
- Tweets4.2K
- Following839
- Followers213
- Likes5.9K
Alexis retweeted
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
gnegnegne no coding needed, toujours vous devez exagérer
Holy moly, DeepSeek Harness can literally modify itself now 😭
They just added Creator Mode, and the idea is kinda insane.
You basically tell it what you want the software to do, and it writes and installs the plugins for you. No coding needed.
Want a video editor? Build one. Need a research paper reader? Add one. You can even create custom tools, workbenches, or redesign the entire UI.
The demo even shows off a sandbox game and a 3D LLM visualizer lol.
What's interesting is that it's built on Cordis, where pretty much everything is a plugin, even the interface itself.
We're getting to a point where instead of waiting for developers to ship a feature, you can just ask the software to build it.
That's a pretty crazy shift when you think about it.
Alexis retweeted
We @huggingface reproduced this paper!!
It's built with ML Intern, an open-source harness to do AI research, train models, and more
Model: huggingface.co/ML-Intern-lab…
Demo: huggingface.co/spaces/ML-Int…
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Tous les jours il y a une dinguerie, mdr je vais voir si je peux tester ce soir
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Alexis retweeted
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Voilà ça c est tout de suite plus interessant
🚨梁圣出招了！DeepSeek Harness 正式上线创作者模式！
一句话就能把想法变成可直接用的插件，从设计工作台、便签墙、倒计时，到沙盒游戏、3D LLM 可视化、视频剪辑台都能生成。
亮点速览：
🔹自然语言生成插件，自动安装可用
🔹基于「Everything is a plugin」架构，真正自定义 Harness
🔹官方放出 7 个演示 Prompt（中英双语）可
Prompt 合集：cdn.deepseek.com/harness/dsh…
你最想造什么插件？
#DeepSeek #DeepSeekHarness #CreatorMode #创作者模式 #AIAgent #一切皆插件 #梁文峰 #崔添翼
teapaw.org/nft_chen/status/211723…
Alexis retweeted
Vous voyez qu'on peut faire du travail de qualité avec de l'IA ?
Ça demande des skills et pas mal de temps d'apprentissage
Mais c'est mieux que les AI Slop qu'on nous pond comme succès nationale ici ptdr
Alexis retweeted
On a fait de petits rajouts sur MakiSend
Si tu ne sais pas c'est quoi c'est un SaaS open source pour envoyer des e-mails transactionels à grande échelle que tu peux deployer en un clic sur ton compte CloudFlare (ça tient entièrement sur le plan free):
- Pas d'infra
- Pas de serveur
- Conçu pour gérer efficacement de gros volumes
Maintenant on a eu pas mal de retours de quelques personnes qui voulaient aussi host ça dans leur infra d'entreprise. On a donc publié une image docker, une image pour run le dashboard et l'API et toutes les features de MakiSend.
Toujours dans la simplicité. Plus de détails sur notre site 👇🏽
mdr les insecure hun, c'est sur twitter j'apprends qu'on a pas le droit de complimenter le mari ou la femme de quelqu un
Alexis retweeted
The NSA built a full reverse engineering suite and gave it away for free. It's called Ghidra.
THE SAME TOOL THE NSA USES TO TAKE APART MALWARE IS OPEN SOURCE. WHAT COULD YOU TAKE APART WITH IT?
github.com/nationalsecuritya…
