Programador de Guadalajara. Me gustan linux, el software libre, el servicio social, el fucho, la bici y la guitarra.

Guadalajara, México
Joined July 2007
Joaquin Bravo retweeted
Big win for local LLM inference! This new inference engine runs LLMs 4x faster than llama.cpp: I ran Qwen3.5 9B on a base M5 MacBook Pro with three engines: - llama.cpp: 22.0 tokens/s - MLX: 25.1 tokens/s - uzu: 92.1 tokens/s The video below shows all three runs side by side. The first two numbers don't mean llama.cpp or MLX are badly built. But both are sitting right under a hard limit of local decoding. When an LLM generates text, every new token needs one full pass through the model, and that pass reads all of the model's weights from memory. A base M5 reads memory at about 153 GB/s. A 4-bit 9B model weighs about 5.2 GB. So at one token per pass, the ceiling is roughly 153 ÷ 5.2 ≈ 29 tokens/s, however fast the GPU computes. uzu gets past that ceiling by producing several tokens per pass instead of one: ↳ A small drafter (DFlash) proposes candidates for the next 16 positions at once. ↳ A tiny adapter (Weaver, 56.7M params) picks among them left to right, so the guesses form coherent text, and keeps a few alternative branches as a tree. ↳ The 9B model checks the whole tree in a single pass and keeps the longest path it agrees with. The main model still decides every token that ends up in the output. GitHub repo: github.com/trymirai/uzu (don't forget to star it 🌟) In my run, each pass of the 9B model produced 7.5 tokens on average. Ordinary decoding produces exactly 1. The weights and kernels are designed around this too. uzu ships its own 4-bit checkpoints, and on M5 chips it runs the verification step on the GPU's int8 matrix path. uzu is an open-source inference engine by Mirai Labs, written in Rust with Metal kernels. It has Swift, Python, TypeScript and Rust bindings, so the same engine runs inside a Mac or iOS app. I shared the GitHub repo above. If you want to dive deeper, my co-founder wrote a detailed breakdown of why local inference is memory-bound and how uzu's speculative decoding gets past it. Read it below.
73
95
14
885
56,130
Joaquin Bravo retweeted
Entire is the ONLY company I've invested in the last two years: b/c it's clear that we need an "AI-native" GitHub, and also clear that GitHub cannot disrupt itself. Entire just reinvented the PR for agents with Trails, and offer a Git forge that can handle agent load:
Agents reinvented how we produce code. It changed our tools and half of our workflow. It’s time for the other half to catch up. Today, we’re shipping Entire: the Git forge for agent-human collaboration. We’ve built @GoEntire using Entire. Now we want to see what you'll build with it. Here’s why and what’s different. 🧵
81
28
3
685
69,412
Joaquin Bravo retweeted
不想传播焦虑,但这次我无话可说。 今年科学界最大的新闻,没有之一。OpenAI 的内部 AI 模型自己做出了三个科学难题的成果。我用我自己的理解说一下。 第一,Kakeya 猜想。这个问题研究波聚焦时会怎样。原来的猜想我记得是说最极端的聚焦只会持续很短时间。人类以前只在低维度证明过,OpenAI 说它的模型证明了所有维度的情况,因此顺带解决了 Kakeya 猜想和 Stein 限制猜想。纯逆天。 第二,指数时间假说 ETH 被推翻。这个太难以置信了。ETH 好像是说有一类逻辑难题不可能被比暴力穷举快很多地解决。而之前学界普遍相信它成立,几百篇论文都以它为前提,现在 AI 却证明它是错的。 这个太他喵逆天了,因为我之前是研究过一段时间理论计算机的。这几十年ETH几乎被当成公理在用。很多论文的结论都写成假设 ETH 成立,那么某某问题不可能在多少时间内解决。现在地基没了,这些不可能,全都要打问号。 最让我感觉匪夷所思的是 AI 证明的方式。据 Lean 证明的说明文档,它给出的是一个确定性算法,应该是对固定宽度的 k-CNF 公式,它能在亚指数时间、多项式空间内精确数出所有满足的赋值。但是,懂行的人应该都知道“数出有多少个解”通常比“判断有没有解”难得多,它连更难的那个都做到了,所以 ETH 只是顺带被推翻。 第三,杨-米尔斯、量子色动力学。AI 应该是给出了一套严格的数学证明,证明描述强核力的理论在数学上确实成立,其中包括纯杨-米尔斯理论的存在性,这是千禧年七大难题之一的一部分。我不太了解,但是只是知道一些浅层的理论,就不评论了。 最让我恐惧和惊骇的是第二个。今天晚上我看了又看,感觉这是地震级别。如果最终确认无误,这会是 AI 第一次独立推翻一个被广泛相信的核心猜想。说它是一个时代的分界点,也毫不夸张。 如果这三个都核查通过,这代表 AI 第一次在多个顶尖领域独立解决了几十年没人能解决的难题,影响史无前例。
Announcing three major scientific discoveries, one each in mathematics (harmonic analysis), computer science (complexity theory), and physics (quantum field theory). AI starting to help us improve our understanding of the universe! openai.com/index/building-sc…
109
335
52
2,585
426,044
Joaquin Bravo retweeted
🏎️ I've never seen MoE models for embedding & search this small: 1.5M, 27M & 115M parameters! @VAGOsolutions' MoEdernColBERT models take on much larger dense retrievers, with 115M near or ahead of Qwen3-Embedding-0.6B in their overview at 8× the encoding throughput. Details 🧵
6
12
3
114
4,245
Joaquin Bravo retweeted
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence. block.xyz/dessa
4,178
7,124
4,731
69,044
13,441,190
Joaquin Bravo retweeted
Sadly canceling our linear subscription as we just don’t use it anymore Planning happens in a Slack canvas with very high level goals and constraints Features gets shipped in a few hours, or days max Roadmaps don’t get further than a month
63
15
27
576
242,120
在屠宰场干了5年,告诉你一个秘密:动物今天要死,它们全都知道 我在屠宰场干了整整五年,每天跟猪牛羊打交道,见过的生死场面成千上万。说句掏心窝子的话,没入这行之前,我跟大伙儿一样,觉得牲口嘛,吃吃睡睡,稀里糊涂一辈子,哪懂什么生死。可五年下来,我摸透了它们每一个眼神、每一个动作,才发现自己当初多天真——这些不会说话的家伙,心里跟明镜似的,自己哪天走,清清楚楚。 平日里,待宰圈那叫一个热闹。猪拱猪、羊顶羊,抢食的打闹的,眼神清亮温顺,该吃吃该喝喝,天塌下来都不耽误长膘。可一到屠宰当天,那气氛说变就变,比翻书还快。整圈牲畜像商量好了似的,瞬间安静下来,挤成一团,紧紧挨着彼此,浑身抖得跟筛糠一样。那抖不是冻的,是打骨头缝里渗出来的恐惧。你赶它走,它四只脚死死扒着地,几百斤的大猪平时一轰就跑,那天愣是钉在原地往后缩,好像前面那条通道是鬼门关。老话讲“万物有灵”,我原先半信半疑,后来彻底信了。 牛的表现更让人心里发酸。但凡当天要走的牛,从早上起就不吃不喝,低着头杵在墙角,一动不动。我亲眼见过无数次,大颗大颗的泪珠子顺着牛眼角往下滚,湿漉漉的眼睛望着你,不挣不扎不吼叫,就那么默默流泪。有一回,一头养了好几年的老牛被主人牵来,主人站栏杆外看了一眼,那牛瞬间红了眼眶,直勾勾盯着主人,温顺又卑微,一步三回头地跟着我们走。那画面,我这辈子忘不掉。羊也一样,平时蹦蹦跳跳软软糯糯,临了却异常安静,紧紧偎着同伴发抖。赶别的羊成群往前冲,唯独那天,你推我挤谁都不肯带头,偶尔被单独牵出来的小羊,发出细细的凄厉叫声,不是闹腾,是真害怕。 时间一长,我们工人不用看单据不用记编号,单凭状态就能判断个八九不离十——活蹦乱跳的是来日方长,沉默发抖、不吃不喝、眼神惶恐的,全是今日宿命。你说它们听不懂人话、看不懂流程,怎么就掐得这么准?其实道理不复杂,流水线一开,前面的同伴一个个消失再没回来,后面的全看在眼里;环境里那股肃杀的气息、同类压抑的悲鸣,它们比人敏感百倍。俗话说“蝼蚁尚且贪生”,何况这些大牲口?它们沉默、发抖、流泪,不是傻,是无力回天的绝望。 五年下来,我这心态彻底翻了个个儿。刚入行那会儿年轻心硬,只当是谋生糊口的活儿,见多了也就麻木了。可干得越久,心越软,越觉得该存一份敬畏。走出屠宰场,我再也不糟蹋粮食,每一块肉背后都是一条热烈活过、恐惧过、挣扎过的命。它们从出生到落幕,没做错任何事,却坦然接受了既定的安排。反观咱们,衣食无忧平安顺遂,倒常常为鸡毛蒜皮纠结、为一点得失内耗,想想真有点滑稽。 所以你看,活着本身就是天大的福气。所有的矫情、焦虑、委屈,搁在生死面前,屁都不是。心存敬畏,懂得珍惜,善待万物,大概就是人这辈子最好的修行。说到这儿我倒想问问你——下次端起饭碗的时候,会不会也想起那些沉默发抖的生灵,然后觉得,眼前这口热乎饭,真香?
1,130
7,668
816
34,364
3,272,421
Joaquin Bravo retweeted
messaging my wife a link containing proof of her husband cheating on her from an anonymous instagram account and then when she clicks on it texting her to let her know she failed the family phishing test and she has to retake my cybersecurity course
906
15,120
1,224
246,082
8,143,522
Joaquin Bravo retweeted
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model. 🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal. 🔹 Full-project coding: games in Unreal, Unity & Godot. 🔹 Excels at parametric 3D modeling in Blender & CAD. 🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%. 🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks. Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API. Open weights by Nov 4, 2026. 🔗 API: platform.kimi.ai 🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
1,249
3,231
2,354
30,137
4,025,014
Joaquin Bravo retweeted
This has been obviously destined since Dario saw GPT-3 autocompleting a for-loop in python and was struck by a vision of our future. My guess is that this has barely started, and that Super Worms are coming. There's never been a better time to build an intranet for your people.
welp. the Internet was nice while it lasted.
59
105
38
2,660
213,696
Joaquin Bravo retweeted
I read this and I'm terrified, the situation seems intractable. In the python ecosystem alone every cpython install, venv, UV is at risk. Don't get my started on npm, pnpm, yarn, go, rust. My current 2026 security posture: stay home and only connect to the internet from a trusted ISP. The winter is coming. discuss.python.org/t/informa…
73
374
68
3,278
400,559
Joaquin Bravo retweeted
Some stats on #php2shell. In the 10 hours before it was publicly announced, across the @Cloudflare network we saw thousands of attempts to exploit it. That jumped to tens of thousands in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now.
289
3,275
673
22,653
2,267,690
Joaquin Bravo retweeted
Guys I’m scared the takeoff is starting. I swear any one of the ~10 CVEs that dropped in the last 24 hours would’ve been “one of the worst in recent years” except a new one is dropping every hour
Vercel has proactively rolled out Firewall-level mitigations for today's Node.js RCE vulnerability, in advance of the official patch release. All traffic for all deployments is automatically protected. We're actively monitoring and rolling out managed runtime upgrades.
80
252
45
4,229
1,693,606
Joaquin Bravo retweeted
NVIDIA just released Flare, a local inference server built for DGX Spark and RTX Blackwell GPUs - Built for long chats, agents and tool calls - Serves from Rust, no Python - Plans memory up front - FP8 KV cache by default Interesting timing... right around when DeepSeek open-sourced their inference engine. Link below.
59
74
16
896
67,977
Joaquin Bravo retweeted
Congrats to @nvidia for creating the best open-weight model on document parsing so far. Nemotron 3.5 Parse takes the #1 spot on ParseBench, our comprehensive real-world benchmark for document understanding. The full list consists of 150+ frontier models, open-weight models, and specialized solutions. It does this while remaining sub-1B parameters so it's also blazing fast ⚡️ Big thanks to @JensenHuang for the shoutout at GTC Berlin! Nemotron release: huggingface.co/blog/nvidia/n… ParseBench: parsebench.ai/
46
106
17
1,006
54,434
Joaquin Bravo retweeted
So basically, today PHP got two critical bugs that let anyone run code on your server without even logging in, which is bad, and then the details leaked before the patch was even out, which is worse, and it works on WordPress and MediaWiki and Magento and Joomla and Laravel, and Censys says like 1.5 million servers are running a version with no fix, so the NSA was like "take your system offline," and the FBI was like "take your system offline," and CISA and the UK and CERT-EU were all like "take your system offline," and then the Prime Minister of Japan, the actual Prime Minister, posted at 5:55 in the morning telling 48.8 million people to take their system offline, so everyone took their system offline, like the Dutch cybersecurity agency took its own website offline, and the Dutch parliament went offline, and the Port of Rotterdam went offline, and Finland and France and Denmark started pulling government sites, and Wikipedia went down, and Archive.org went down, and Asana and GOG and Something Awful went down, and the University of Vienna's webmail went down for some reason, and that's just PHP. Because GitLab also dropped a CVSS 10.0, which is the highest score they give out, for a bug that's been there since 2014, so that's every self-hosted GitLab for twelve years, and then LiteLLM 1.107.0 got hijacked and shipped a credential stealer and a remote access trojan in every single wheel and the Docker images, including latest, and you didn't even have to import it, just installing it was enough, and it's the second time this year, and then Railway paused every deployment on the entire platform because someone got into the environment variables, so now everyone has to rotate every secret they've ever stored there, and TiDB Cloud said someone was walking around inside with stolen credentials for basically a month and an extortion group says they took 10.6 terabytes, and Zeabur got breached for the second time in two months and is just shutting down, like the whole company, and Node.js patched remote code execution and a TLS bypass, and CISA says Plex is being actively exploited, and someone took over 300 Tor exit relays to steal SSH passwords, and SafePay ransomware has hit over 600 critical infrastructure orgs, and India's biggest bank filed a data loss incident, and Japan has been in an official state of emergency in cyberspace since October 9, which honestly explains the 5:55 AM post. Oh, and eMudhra, a certificate authority that's issued over 60 million certificates, whose entire job is proving websites are who they say they are, was supposed to be using private keys that are random numbers over a hundred digits long, and one of them was 4. Just 4. And the others were like 66 and 237. Anyway, it's Wednesday.
1,078
8,935
1,652
69,411
4,183,860
Joaquin Bravo retweeted
🚨 @NVIDIAAI just quietly dropped its own local inference server... No announcement. Single-digit stars It's called Flare and Apache 2.0 Built for SM120/SM121 GPUs RTX PRO 6000 Blackwell and DGX Spark >PyTorch + FlashInfer at build time >Rust runtime at serve time. No Python. >CUDA graphs captured before serving >FP8 KV cache by default >Recipes for Qwen Flash, GLM and DeepSeek github.com/NVIDIA/flare-loca…
48
86
27
740
68,844