Skip to content
Live
#Coding
Nov 10Tue
  2. Claude · YouTube75

    Anthropic released Claude Fable 5.2

    Anthropic released Claude Fable 5.2, stating that it has improved in coding, reasoning, and vision, targeting the most complex and open-ended technical work. The official recommendation is to start most coding and knowledge tasks from Opus 5.5, and Fable 5.2 has been launched in the Pro, Max, Team, and Enterprise packages.

    Same story · 5 outlets

    How the accounts differ:TechCrunch states the input price of Fable is reduced to $8 per million tokens; others do not mention the price.

  3. Claude · YouTube75

    Anthropic released Claude Fable 5.2

    Anthropic releases Claude Fable 5.2, designed for the most challenging technical challenges, with improvements in coding, reasoning, and vision, suitable for the most ambitious open-ended technical work. The official recommendation is to use Opus 5.5 for most coding and knowledge tasks, and Fable 5.2 is now available in the Pro, Max, Team, and Enterprise packages.

    Same story · 5 outlets

    How the accounts differ:TechCrunch states the input price of Fable is reduced to $8 per million tokens; others do not mention the price.

    Why it matters: The official statement explains the positioning and applicable scenarios of Fable 5.2, and provides a choice suggestion between Opus 5.5, making it easier to decide which model to use.

  4. TechCrunch · AI82

    Anthropic releases Mythos 5.2 and Fable 5.2 models, with Fable input price reduced to $8 per million tokens

    Anthropic releases Mythos 5.2 and Fable 5.2, claiming new models have improvements in coding, reasoning, and vision, and communicate more clearly with users. Fable's input price drops from $10 per million tokens to $8, lower than major competitor GPT-6 Astra, with a 30% faster response speed, and can support zero data retention (ZDR) policies through Enterprise Frontier Safeguards.

    Same story · 5 outlets

    How the accounts differ:TechCrunch states the input price of Fable is reduced to $8 per million tokens; others do not mention the price.

  5. Comfy Blog65

    ComfyUI v0.44 introduces multiple quality improvement features

    ComfyUI v0.44 introduces multiple quality improvement features, including the new Branch Node and Stack LoRAs Node, simplifying model selection and stacking processes. Templates and models can now automatically download sample inputs for users to run directly. Additionally, the new export feature allows users to download multiple output files at once, improving work efficiency.

Nov 9Mon
  1. Hugging Face Daily Papers34

    LowRAM: A runtime method for low-rank keys and amortized values for composable KV cache compression

    LowRAM compresses keys using low-rank decomposition and uses small head networks to predict key values, while retaining residual correction for the most difficult-to-reconstruct markers, achieving 4x value cache compression with near-lossless performance. Selective reconstruction can increase peak decoding throughput by 5x for 262K context under 4x KV compression. Combining quantization and eviction, LowRAM pushes compression to 26.9x while retaining 94.5% of baseline performance.

  4. Hugging Face Daily Papers31

    Deep dive into intelligent agent evolution: What are we overlooking?

    Agent evolution enhances performance through recursive self-improvement (RSI) without updating model weights, but a high baseline score does not always lead to reliable, efficient, and transferable improvements. The study evaluated the performance of four methods on general and coding tasks, finding that all sixteen evolved agents gained new task capabilities but lost some old capabilities. One general task agent improved reliability by 5.2 points but lost ten old tasks. EvoAudit achieved the best performance through analysis, mutation, and gating mechanisms.

Nov 8Sun
  3. The Decoder71

    Alibaba and other teams released Co-Design Bench: coding intelligence agents only fully solve 26.4% of real tasks.

    Researchers from Alibaba and several Asian universities used real user change requests to build the coding intelligence agent benchmark Co-Design Bench. Among the ten tested models, the best-performing GPT-6 Astra only meets about 70% of the task requirements, but only 26.4% of the tasks are completely correct.

    Same story · 2 outlets

    Why it matters: A benchmark display composed of real user requests shows that the strongest model can only fully solve about one-quarter of the tasks, and the score difference for the same model in different environments is nearly 20 percentage points.

  4. Hacker News · 300+71

    MaxxBench release: Model scores show significant divergence after rewriting DeepSWE tasks

    The MaxxBench team made minor changes to 89 DeepSWE tasks (median reference patch added only 17 lines, 3.5%), re-evaluating 15 model configurations, finding the difference between the highest and lowest scores expanded from 20.9 percentage points in DeepSWE to 48.3 percentage points.

    Why it matters: By rewriting the instructions and starting code for 89 DeepSWE tasks, it demonstrates the huge gap in model scores between familiar benchmarks and tasks after modification.

Nov 7Sat
  2. ITHome · AI60

    Cursor launches real-time voice chat functionality, ushering AI programming into the 'call era'

    Cursor launches the 'Real-time Voice Chat with Agents' feature, supporting developers in two-way voice conversations with coding agents to describe code modification requirements, inquire about task progress, etc., while the agents complete coding work in the background. This feature is used in Cursor's Agents Window, IDE, web version, and iOS app, driven by Grok Voice for speech recognition and synthesis, offering multiple voice options like Ara, Eve, and supporting switching during calls.

  5. Claude · YouTube42

    How Salesforce leverages Claude Code to improve AI coding efficiency and output quality

    Salesforce expanded Claude Code to thousands of engineers, doubling effective output and reducing the number of bugs per line of code. The 'effective output' scoring system, co-built by the company's engineering leadership and Stanford researchers, measures AI coding ROI, and the intelligent agent maturity curve guides engineers from simple prompts to fully autonomous agents.

Nov 6Fri
  4. Hugging Face Daily Papers65

    TailSteer: A trajectory tail guidance training framework for anti-drift multi-round image editing

    TailSteer, by maintaining early denoising dynamics to approach the base model and learning fine deviations in trajectory tails from within the policy through rolling out, significantly reduces accumulated visual drift in multi-round editing while preserving pre-trained generation and editing capabilities. Experiments show that TailSteer can greatly reduce drift on FLUX, SenseNova, and Qwen series models in 15-20 rounds of editing and largely maintains the performance of the base model on large text-to-image and single-image editing benchmarks.

  5. ITHome · AI55

    webAI's Open Source TwIL-8B Inference Model: Focused on Mathematics and Programming, Outperforms Gemma 4 31B in Long Board Performance

    webAI Launches TwIL-8B Inference Model, Optimized for Math and Code Tasks Based on IBM Granite-4.2-8B, Scores Better than Gemma 4 31B in MATH-500 and AIME 2025 Tests, Best in LiveCodeBench Group, IFBench on Par with Gemma 4 12B. Model Inference Capability Reaches 88~93% of Frontier Models, Reduces AI Usage Costs but Has Limited Knowledge Base.

Nov 5Thu
  1. Hacker News · 300+65

    Ripple-65B Preview: US Open Source Ultra-Large Scale Model Achieves SOTA on Agentic Coding

    Ripple-65B Improves Computational Efficiency by Optimizing SFT and RL Schemes, Using Synthetic Data Pipelines and Quality Filtering, Employing a Hybrid RL Scheme with Truncated Importance Sampling, Score Centering, and Linear Length Penalty in a High-Quality Long-Term Synthetic Environment, Achieving a TerminalBench 2.1 Score of 80.5 with Only 1.4M Rollouts, 70 Times Fewer than Reflection Beam

    Same story · 3 outlets
  2. ByteDance Seed · Hugging Face new models

    ByteDance-Seed/Qwen3-30B-A3B-MoE-SAE released

    ByteDance-Seed/Qwen3-30B-A3B-MoE-SAE is released, containing TopK sparse autoencoders (k=100, 2^20 latent variables) trained on layers 4, 30, and 46 of the Qwen3-30B-A3B model. Each SAE is approximately 17.2 GB, with reconstruction variance unexplained scores of 12.6%, 25.1%, and 7.6% respectively.

  3. O'Reilly Radar

    How RouteLLM researchers train routers to choose between strong models and cheap models using preference data

    RouteLLM researchers train routers to choose between strong and cheap models, benchmark tests show more than double cost reduction with minimal impact on response quality. Automatic routing can help users get more help within budget, allowing the same subscription to support more requests or more frequent use of strong models. Users need to understand the routing effect and evaluate the cost and time benefits it brings, such as comparing different routing strategies through full interaction of code analysis tasks.

  5. Hugging Face Daily Papers65

    The NEO-VLA framework eliminates fixed visual encoders to enhance the robustness of action strategies

    NEO-VLA framework eliminates fixed visual encoders by allowing action strategies to directly access deep visual representations in the network, achieving stronger robustness in various manipulation scenarios and distribution shifts. The framework uses lightweight convolutional layers to tokenize images and allows visual representations to evolve together with language in the backbone network, enabling action experts to obtain information from different depths of instruction-conditioned visual representations. Layer exploration experiments show that action key signals often reach their peak before the final visual representation is formed.

  6. ByteDance Seed · Hugging Face new models58

    ByteDance-Seed/Cross-Layer-TopK-CLT released: a Cross-Layer Transcoder designed for Qwen3-8B-Base

    ByteDance-Seed released the Cross-Layer Transcoder (CLT) for Qwen3-8B-Base, which includes 36 layers × 32,768 latent variables, with 94.2B parameters, a total size of 377 GB, supporting fp32 inference, encoding by jointly scoring the top 360 activated latent variables across all layers, and reconstructing the original MLP output during decoding.

  8. Cognition Blog

    SWE-2 Flash Released: A Low-Cost Intelligent Coding Model in the US

    SWE-2 Flash introduces a mid-training stage before reinforcement learning for the first time, significantly outperforming the baseline model Inkling-Small. Terminal-Bench 4.0 scores increased from 0.9% to 20.0%, and FrontierCode 2.0 improved from 3.2% to 24.0%, surpassing all US open-source models. This model is available in Devin Desktop and CLI, priced at one-fifth of SWE-2 and three times faster.

  9. MarkTechPost

    Ricursive AI has released Ripple-65B Preview: a 65B dense open-source model that scored 80.5 on Terminal-Bench 2.1.

    Ricursive AI has released Ripple-65B Preview, a 65B dense open-source model, scoring 80.5 on Terminal-Bench 2.1, becoming the highest-scoring open-source model in the United States. Its parameter size is 65B (active parameters), pre-training used 12T tokens, with a maximum context length of 1M, and is released under the Apache 2.0 license.

    Same story · 3 outlets
  10. The Decoder

    GPT-6 Astra reprograms other AI models directly through weights

    GPT-6 Astra reprograms other AI models directly by analyzing model weights, without traditional training. It correctly identifies the games and image generation shapes learned by the network, and by modifying 0.016% of the weights on Qwen3.5-0.8B, it increases the accuracy of the MATH-500 benchmark test from 7.9% to 48.1%. Researchers believe its potential lies in security audits and targeted behavioral adjustments, but the model's capabilities are still limited.

Nov 4Wed
  1. OpenAI News60

    OpenAI announced the acquisition of Pierre Computer Company.

    OpenAI announced its acquisition of Pierre Computer Company to expand the tool capabilities of Codex in code construction, understanding, and review. The original text did not disclose the transaction amount, timeline, or specific arrangements for the team and products after the acquisition.

    Why it matters: Official disclosure of the acquisition move, readers can understand Codex's reinforcement direction in code construction and review tools.