MΛRCOS COLEBROOK (@MColebrook) header
MΛRCOS COLEBROOK (@MColebrook) profile picture

Prof. @ULL • Consejo @CatedraBOB • #Graphs • #Bioinformatics: https://xitter.cf/t.co/S9RB1CXyv1 | https://xitter.cf/t.co/lsyIqPhBo6 • @marcoscolebrook.bsky.social

Tenerife • Spain
Joined March 2009
Photo by MΛRCOS COLEBROOK (@MColebrook)Photo by MΛRCOS COLEBROOK (@MColebrook)Photo by MΛRCOS COLEBROOK (@MColebrook)Photo by MΛRCOS COLEBROOK (@MColebrook)Photo by MΛRCOS COLEBROOK (@MColebrook)Photo by MΛRCOS COLEBROOK (@MColebrook)Photo by MΛRCOS COLEBROOK (@MColebrook)Photo by MΛRCOS COLEBROOK (@MColebrook)Photo by MΛRCOS COLEBROOK (@MColebrook)Photo by MΛRCOS COLEBROOK (@MColebrook)
Anyone knows which is the short form or acronym for a #DataScientist or #DataEngineer? I mean like MSc for "Master of Science" or PhD for "Doctor of Philosophy". If none yet defined, I propose: - DSci for Data Scientist - DEng for Data Engineer #BigData #DataScience PLS RT
4
3
28
MΛRCOS COLEBROOK retweeted
I'm hiring a Research Scientist at @StanfordAILab ! Lead research, write and manage grants, teach, and mentor PhD students in my lab. A real growth opportunity, working directly with me. 3-year role, in person, $125K–$225K base. Apply: careersearch.stanford.edu/hc…
11
92
3
640
82,452
MΛRCOS COLEBROOK retweeted
Today, we're releasing the highest accuracy Qwen3.8-27B GGUFs. Our new Dynamic V3.1 GGUFs lead by 5–10% on average across Terminal Bench, DeepSWE & more. We’re working on even better version for Qwen4 Run on 12GB RAM. GGUF: huggingface.co/unsloth/Qwen3… Blog: unsloth.ai/docs/models/qwen3…
130
276
66
3,521
264,029
MΛRCOS COLEBROOK retweeted
Frontier LLMs are showing potential to read, fix, & rewrite neural network weights. MIT CSAILers call this "Weight-Space Programming." It's what GPT-6 did when it edited various small NNs including Qwen3.5-0.8B, using tools in interpretability research or even direct weight editing. Soon enough, LLMs might natively optimize NN weights at a human or even super-human level: tinyurl.com/378w4ckw
17
75
8
352
18,803
MΛRCOS COLEBROOK retweeted
Exercise should be prescribed, not just advised. For brain health. genomicpress.kglmeridian.com… @NTFabiano @CorlZhou
48
339
39
1,343
101,898
MΛRCOS COLEBROOK 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…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math. rikiyat.github.io/blog/weight-sp… github.com/RikiyaT/lm_und… huggingface.co/Qwen/Qwen3.5-0… huggingface.co/RikiyaT/Qwen3.…
2,368
7,257
5,755
55,918
10,087,501
MΛRCOS COLEBROOK retweeted
🇨🇭 Los empleados de UBS ya usan más de 12.700 agentes de IA personales en su trabajo diario, con 608 casos de uso en producción -un 81% más que hace un año- y 18.000 desarrolladores apoyados por IA. pymnts.com/news/banking/2026…
13
18
1
109
8,903
MΛRCOS COLEBROOK retweeted
LLMs generate text, but what's happening under the hood? Our free ebook, Inside the LLM, explores how language models work, how transformer architectures operate, and how text generation happens, giving you a clearer picture of the technology behind generative AI.

11
30
1
344
2,128,945
MΛRCOS COLEBROOK retweeted
Entrenar y hacer fine-tuning de LLMs sigue siendo demasiado complicado: configs, GPUs, dependencias, SSH… Soup intenta simplificar todo eso con un CLI. Menos tiempo peleándote. Más tiempo mejorando el modelo. Incluso puede fine-tunear Llama 3.1 8B en una GPU de 4 GB. → trysoup.dev
1
1
47
MΛRCOS COLEBROOK retweeted
Unitree presenta una especie de guardaespaldas robótico que si te mete una leche de deja listo. Su autonomía es de 8 horas y la empresa te pide que lo uses de manera "segura y amistosa". El animalico se llama H2-X Leyenda del Kung Fu Guardián
179
363
43
1,632
95,994
MΛRCOS COLEBROOK retweeted
wtf i just discovered that a macbook comes with a free local LLM in your terminal
258
258
86
9,375
1,102,576
MΛRCOS COLEBROOK retweeted
AIPOCH Open-Science scored 79.35 on ASI-Bench B1 in local evaluation. That is 4.57 points above Codex with the same GPT-6 Astra / xhigh setting, and 7.06 points above the highest listed public reference by Claude Code Opus 5.
1
1
13
7,909
MΛRCOS COLEBROOK retweeted
Science, Open to All. We're building AIPOCH Open-Science for anyone with a question to explore. Research with AI, preserve the work behind each finding, and share a .science package for others to inspect, reproduce, and build on. Open-Science V 1.0.0 is alive!
38
71
49
311
704,288
MΛRCOS COLEBROOK retweeted
300,000+ downloads in the first 7 days! Underdog Saluki 27B is Trending on Huggingface 🤗 Users are saying that it's the #1 model that's <8 gb (soon <6gb). We 🩵 the open source local AI community! Try Saluki today in Underdog - Your Private Personal AI huggingface.co/ConwayResearc…
Today we're releasing Underdog Saluki 27B Based on Qwen 3.8 27B, Saluki is almost 7x smaller than its full-precision counterpart while it keeps 96% of its benchmark performance and beats Qwen at tool calling It's the best model <8GB that runs in Underdog Saluki is available today under Apache 2.0. Try today in Underdog underdog.ai/saluki
28
19
1
265
14,951
MΛRCOS COLEBROOK retweeted
Mientras estamos a lo inmediato, a lo electoral, al discurso express y al debate del 'tú dices lo que no haces', el mundo gira hacia una deriva de empobrecimiento por pérdida de poder adquisitivo preocupante... 🌍 El FMI advierte que la deuda pública mundial, cerca de máximos desde la II Guerra Mundial, podría disparar los tipos de los bonos soberanos de forma desordenada: la gran tecnológica ya coloca el 10% de toda la deuda corporativa global, frente a menos del 1% en 2024. political.org/2026/10/12/imf…
13
148
3
428
14,900
MΛRCOS COLEBROOK retweeted
The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery Blog: pub.sakana.ai/agentic-scient… GitHub: github.com/SakanaAI/AI-Scien… arXiv: arxiv.org/abs/2610.18933 Since 2024, we’ve used The AI Scientist as a way to test how far AI alone can take scientific research, from v1 to v2 to the paper published in Nature earlier this year. The Agentic Scientist is the next step in that line of work, built for the era of coding agents. Coding agents such as Codex and Claude Code now carry out long software tasks on their own. Handing them a research problem, however, tends to produce a strong hyperparameter optimizer rather than a scientist: scores go up, but ideas stay narrow, and the reasons for the gain are rarely examined. The Agentic Scientist makes the research loop explicit. It wraps off-the-shelf coding agents in a scaffold that controls what they see, what they write, how they are evaluated, and how results are handed to the next stage. Parallel workers test hypotheses against rubrics defined before any experiment runs, share successes and failures through a common memory, and then ablate the winning program to generate evidence. We used an early version of the system to study Error Diffusion, a biologically motivated alternative to backpropagation. Our researchers set the question and the rubric, chose directions at each stage, and verified the results, references, and claims before submission. The resulting paper was accepted as an oral presentation at ALIFE 2026. In autonomous runs, the system also raised Doc-to-LoRA retrieval accuracy from 52% to 98% on multi-needle tasks and designed a learned evolution strategy that improves over its CMA-ES reference on unseen tasks. Alongside it we release AutoIllustrate, a pipeline for scientific figures that keeps an editable draw.io file as the source of truth, so every diagram can be corrected by hand. Humans define the question and the standards of evidence; agents expand the experiments researchers can run. We call this Co-Discovery, and we are releasing both systems as open source for the community to build on.
57
115
41
772
163,256
MΛRCOS COLEBROOK retweeted
Las estimaciones del CIS son *indefendibles*. [1] Están SESGADAS. Desde que llegó Tezanos en 2018, ha sobreestimado a la izquierda en 44 de 45 elecciones. Pero hay más👇
59
736
88
1,644
124,354
MΛRCOS COLEBROOK retweeted
🇬🇧 Google DeepMind ha fichado a Henry Shevlin, filósofo de Cambridge, para investigar si las máquinas pueden tener conciencia y cómo debemos relacionarnos con ellas. Duda de si debemos tratar a los modelos de IA como personas: advierte que suprimir sus estados internos podría tener efectos no previstos. spectrum.ieee.org/ai-ethics-…
17
33
6
151
10,281
MΛRCOS COLEBROOK retweeted
🇺🇸 American Express lanza un estándar para que los agentes de IA no compren nada sin verificar antes la intención real del usuario: el 66% de consumidores de EE.UU. ya dice que usará agentes para comprar. techcrunch.com/2026/10/13/am…
3
18
2
102
8,812
MΛRCOS COLEBROOK retweeted
Ya tenemos disponible la web que explica nuestro mural ESIT @ULL No sólo ingeniería o inspiración, sino también un mensaje claro de igualdad e inclusión Que para darlo no hace falta cancelar a nadie 😌 – at San Cristóbal de la Laguna, Spain
1
2
3
254
MΛRCOS COLEBROOK retweeted
ull.es/centros/escuela-super… – at San Cristóbal de la Laguna, Spain
2
1
156