Alexander Stojanovic (@stojanovic) header
Alexander Stojanovic (@stojanovic) profile picture

Synthetic Biology, CRISPR, GPGPU, Japanese Design, Electronica, Game Theory, AI, Contemporary Art+Cinema, and Leica Photography [Views expressed are my own]

Los Gatos, CA
Joined April 2007
Photo by Alexander Stojanovic (@stojanovic)Photo by Alexander Stojanovic (@stojanovic)Photo by Alexander Stojanovic (@stojanovic)Photo by Alexander Stojanovic (@stojanovic)Photo by Alexander Stojanovic (@stojanovic)Photo by Alexander Stojanovic (@stojanovic)Photo by Alexander Stojanovic (@stojanovic)Photo by Alexander Stojanovic (@stojanovic)Photo by Alexander Stojanovic (@stojanovic)Photo by Alexander Stojanovic (@stojanovic)
Monoculture B&W Photo Editor from Future Pixels Studio, optimized for Leica Q(tm) and Leica M(tm) cameras. Note: not affiliated with Leitz Camera AG.
2
66
I am thrilled to have my Leica-optimized B&W photo editor Monoculture is generally available in the AppStore (iOS, macOS). If you long for an elegant workflow and UX, zero subscriptions, and no recurring costs, check out Monoculture. apps.apple.com/us/app/monocu…
1
5
177
Alexander Stojanovic retweeted
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected. Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions. Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent. But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models? We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem. Read the full report: cohere.com/blog/concerning-b…
974
1,283
1,522
11,087
3,106,022
Alexander Stojanovic retweeted
Kimi K3.1 Pro launches at 4th on the Artificial Analysis Intelligence Index scoring 55, making large gains in agentic performance; it will be the leading open-weights model once its weights are published @Kimi_Moonshot's new reasoning model gains 11 points over Kimi K3 (max, 44). It trails only Claude Opus 5.5 (max with fallback, 58), Claude Sonnet 5.5 (max with fallback, 56), and Muse Wonder Full (max, 56) on the Intelligence Index. Once the weights for K3.1 Pro are released it will be the highest-scoring open weights model and the gap between the leading open weights and proprietary models on the Intelligence Index will narrow from 12 points to 3. Key results: ➤ Large gains in agentic performance: Kimi K3.1 Pro scores 3rd on GDPval-AA with 1,846 Elo, up 312 Elo from Kimi K3, behind only Claude Opus 5.5 (1,866 Elo) and Muse Wonder Full (1,857 Elo). Kimi K3.1 Pro also achieves the highest score of any model on Terminal-Bench 4.0. K3.1 Pro scores 67%, up from 13% for Kimi K3 and ahead of Claude Sonnet 5.5 (64%). AutomationBench-AA, which tests agentic SaaS workflows, rises from 58% to 73%. ➤ Gains in other evals are more moderate: Humanity's Last Exam rises from 47% to 52%, CritPt from 23% to 25%, and SciCode from 59% to 61%. AA-Omniscience Index rises from 20 to 30, as the hallucination rate falls from 53% to 44% and accuracy rises from 48% to 51%. ➤ Cost per Task roughly equal to Kimi K3 at $1.98 per Task: K3.1 Pro uses 55K output tokens per Intelligence Index task, 14% more than Kimi K3 at 48K, while its first-party token prices are 33% lower. Additional model details: ➤ Context window: 1M tokens. ➤ Size: 2.8T total, 104B active parameters ➤ Input modalities: Text and image ➤ Weights and license: Weights currently not published, Moonshot is expected to release the weights soon after launch ➤ Pricing: $2/$10/$0.20 per 1M input/output/cache hit tokens, 33% lower than Kimi K3 on all three rates. Congratulations to the team at @Kimi_Moonshot on the release!
220
486
341
5,474
671,732
Alexander Stojanovic retweeted
Replying to @karpathy
14
75
22
1,091
181,670
Alexander Stojanovic retweeted
All-time best quote from Dune: "Once men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them."
29
167
14
1,239
70,275
Bi Gan and Bertrand Bonello have produced brilliant films about the erosion of human life in the Post-Singularity/AI Era. Resurrection was an amazing Buddhist parable and The Beast used a Henry James novella as a framing to show loss of human being in a world of “AI alignment”.
4
193
Researchers at @ucsc have now reported on what they suggest is the first-ever method for analyzing #RNAsequencing data genome-wide using a “pantranscriptome,” which combines a transcriptome and a pangenome. Learn more: ow.ly/4xYb50Mvl7u
3
12
2,747
Great line from Bret Easton Ellis’s new book. It’s my favorite since Glamorama and is #meta is so many ways. #TheShards #BEE
1
3
610
Bret Easton Ellis truly is his own best literary creation. I am thoroughly enjoying #TheShards as a meta-fictional conceit. This extends the self-reflexive tropes of “Lunar Park” well. "The Shards: A novel" by Bret Easton Ellis a.co/7VYiOcl
290
Metastatic Colorectal Cancers in Mice Virtually Eliminated Using Immunotherapy and Experimental Drug Combo. Learn more: ow.ly/4Oeh50M73aV
5
5
2,236
OpenELM Release by #CarperAI: This is a really interesting intersection of evolutionary programming and large language models. carper.ai/openelm-release/
152
Alexander Stojanovic retweeted
Can neural networks learn to perform explanatory & counterfactual reasoning? Researchers find that an object-centric transformer substantially outperforms leading neuro-symbolic models on two reasoning tasks thought to be challenging for deep neural nets: https://xitter.gq/t.co/IchQKRDjRL
9
159
14
774