DeepSeek Harness is a desktop application for macOS and Windows that processes documents, spreadsheets, and code while scheduling tasks through composable plugins. ·
4 points and 2 comments on Hacker News, linking deepseek.com.
Researchers propose Hierarchical Continuous Diffusion Language Models, which couple discrete token generation with a continuous latent trajectory to improve structured reasoning and mathematical planning tasks. ·
Researchers propose Hierarchical Continuous Diffusion Language Models, which couple discrete token generation with a continuous latent trajectory to improve structured reasoning and mathematical planning tasks. ·
DeepSeek Harness is a desktop application for macOS and Windows that processes documents, spreadsheets, and code while scheduling tasks through composable plugins. ·
Researchers propose Adaptive Reward Routing, a method that dynamically adjusts update locations and reward coordination during the training of joint audio-video diffusion models. ·
Researchers propose Hierarchical Continuous Diffusion Language Models, which couple discrete token generation with a continuous latent trajectory to improve structured reasoning and mathematical planning tasks. ·
Researchers propose Adaptive Reward Routing, a method that dynamically adjusts update locations and reward coordination during the training of joint audio-video diffusion models. ·
Researchers propose Adaptive Reward Routing, a method that dynamically adjusts update locations and reward coordination during the training of joint audio-video diffusion models. ·
Researchers introduce GraphForge, a framework that synthesizes training data for working agents using real files and evidence graphs. Fine-tuning Qwen3.6-27B on 2,169 trajectories improves its GDPVal score to 1445.7. ·
Researchers propose RASO, a framework that retrieves existing agent skills to initialize and refine new ones, reducing reliance on expensive rollouts. ·
Researchers introduce PoS, an inference-time framework that constructs explicit belief states to help large language model agents maintain coherent understanding and detect when they stop making meaningful progress toward goals. ·
Researchers propose Adaptive Reward Routing, a method that dynamically adjusts update locations and reward coordination during the training of joint audio-video diffusion models. ·
Researchers introduce GraphForge, a framework that synthesizes training data for working agents using real files and evidence graphs. Fine-tuning Qwen3.6-27B on 2,169 trajectories improves its GDPVal score to 1445.7. ·
Researchers propose RASO, a framework that retrieves existing agent skills to initialize and refine new ones, reducing reliance on expensive rollouts. ·
HF Daily Papers4h ago▲arxiv.org
X voices
5 stories
X posts other sources are linking to, then named voices on Bluesky and Mastodon
OpenAI releases Codex version 0.160.0, which adds keyboard-accessible browsing for older tasks and middle-click pasting in fullscreen mode on Linux terminals. ·
Unsloth releases a desktop update adding a command palette, shareable settings, and faster Laya decisions. It also reduces peak memory usage by 45% for NVFP4 LoRA training. ·
Unsloth releases a desktop update adding a command palette, shareable settings, and faster Laya decisions. It also reduces peak memory usage by 45% for NVFP4 LoRA training. ·
OpenAI releases Codex version 0.160.0, which adds keyboard-accessible browsing for older tasks and middle-click pasting in fullscreen mode on Linux terminals. ·
GitHub Releases8h ago▲github.com
Infrastructure
5 stories
Data centers, power, chips and what tokens cost
Data-center and energy stories per day
Table
Sat 26
1
Sun 27
1
Mon 28
20
Tue 29
33
Wed 30
35
Thu 01
25
Fri 02
0
Stories from DCD, Data Center Knowledge, EIA and Hacker News, by publication day.
Median output price of models added in the last 30 days, $ per million tokens
One point per OpenRouter reading, last 8 days. Data by OpenRouter, CC BY 4.0.
Output price, $ per million tokens (newest models)
Table
pareto-26.10-preview
3.2
gpt-6.1-sol-pro
10
gpt-6.1-sol
10
claude-sonnet-5.5
10
claude-sonnet-5.5:batch
5
median, all paid models
2
437 paid models listed; median of the 56 added in 30 days: $4.00. Rankings data by OpenRouter, CC BY 4.0.
The European Space Agency selects OVHcloud and CGI to build Digital EO, a system that supplies Earth observation data to scientists, public bodies, businesses, and citizens. ·