Alphabell.

AI makes AI better.

A daily hub for work where AI is part of the loop that improves AI, each entry with its loop spelled out. Read more →

1/4
A loop in motion
AlphaEvolve
Project · Google DeepMind
Gemini models propose code changes
From the labalphaBell Qwen Open27B post-training research, with capability tradeoffs measured.Explore the model → From the labModel ObservatoryMeasured Flint and Qwen traces with published decoder maps.Open the lab tool → From the labMiniMax Video BudgetVisualize sampled frames, two-frame groups and vision placeholders.Try the inspector → From the labMOSS Blind Layout CheckCompare exact text and placement in two local images.Try the blind check → From the labAgent Request Evidence LadderSee what a trace proves about an agent's request and its outcome.Explore the cases →

Today's radar

20 Nov 2026Full radar →

7 picked from 120 candidates · 30 sources read · 80 loops indexed

  1. 01
    Loop of the dayPaperarXiv19 Nov 2026Hardware and chips

    In Situ Training of Optical Neural Networks by Coding Agents

    The loop: A coding agent trains an optical neural network directly from camera measurements by forming a physical understanding of the optics. This tunes the hardware elements, making the resulting optical network better at classification and image encoding.

    Applying coding agents to directly tune optical hardware from camera measurements bypasses the need for expert-built digital models and gradient-based training.

    loop fit 9/10Ilker Oguz, Colin Christian, Christophe Moser, Demetri Psaltis, Gordon Wetzsteinvia arXiv
    In Situ Training of Optical Neural Networks by Coding Agents
  2. 02
    PaperarXiv20 Nov 2026Training data

    AutoMaster: Autonomous Post-Training by Making the Same Model a Better Teacher

    The loop: An LLM agent analyzes teacher trajectories to design skills that suppress undesirable habits and elicit effective behaviors. This improves the teacher model, which then generates higher quality data to fine-tune a better successor.

    loop fit 10/10Guoxin Chen, Shuang Sun, Fanzhe Meng, Xinyu Tang, Ruihua Song, Wayne Xin Zhaovia arXiv
    AutoMaster: Autonomous Post-Training by Making the Same Model a Better Teacher
  3. 03
    Blog postOpenAI19 Nov 2026Tools and harness

    How OpenAI's Performance Factory Keeps ChatGPT Fast

    The loop: Codex-powered workflows automatically find and fix performance issues within ChatGPT. This improves the system's execution speed, which makes the models themselves faster and more efficient at running future workflows.

    loop fit 9/10via OpenAI
    How OpenAI's Performance Factory Keeps ChatGPT Fast
  4. 04
    ProjectGitHub18 Nov 2026Kernels and compilers

    baidu-baige/KODA

    The loop: An open-weight agent uses compiler, correctness, and profiling feedback to optimize CUDA kernels. This improves the execution efficiency of the underlying hardware, which accelerates the agent's own future operations.

    loop fit 8/10baidu-baigevia GitHub
    baidu-baige/KODA
  5. 05
    PaperarXiv20 Nov 2026Self-modifying agents

    Designer-RSI: Evolving Procedural Memory for Graphic Design Agent

    The loop: A graphic design agent evolves a modular skill bank from its own unlabeled design trajectories by contrasting successful and failed runs. This improves the agent's tool execution, making the same system better at fulfilling complex creative briefs.

    loop fit 9/10Hongyang Du, Lan Yan, Christian Flores, Asim Kadavvia arXiv
    Designer-RSI: Evolving Procedural Memory for Graphic Design Agent
  6. 06
    ProjectGitHub18 Nov 2026Self-modifying agents

    chenkezhen480/evolution-computer-use

    The loop: A computer-use agent learns and repairs its own procedures through interaction. This improves its library of executable skills, making the agent better at navigating desktop and browser environments.

    loop fit 7/10chenkezhen480via GitHub
    chenkezhen480/evolution-computer-use
  7. 07
    PaperarXiv20 Nov 2026Self-modifying agents

    How Much Should an Agent Remember? Deployment Lessons from Self-Distilled Memory for LLM Agents

    The loop: A pipeline mines lessons from an agent's own history and consolidates them into a smaller set of guidelines. This improves the agent's procedural memory, making the same system better at completing long-horizon tasks.

    loop fit 9/10K. R. Jayaram, Vatche Isahagian, Aswath Vaithinathan Aravindan, Vinod...via arXiv
    How Much Should an Agent Remember? Deployment Lessons from Self-Distilled Memory for LLM Agents
  8. 08
    PaperarXiv19 Nov 2026Self-reward and self-play

    SWE-growth: Teacher-Guided Adaptive Self-Play for Software Engineering Agents

    The loop: A student coding model attempts repository-level repair tasks, and a teacher model identifies learning opportunities from the failures. These verified outcomes are used as training data for supervised fine-tuning and reinforcement learning, making the student model better at software engineering.

    loop fit 10/10Songshi Liang, Zixuan Huang, Mingxuan Song, Haoyu Tang, Shusheng Xu, Tongkai...via arXiv
    SWE-growth: Teacher-Guided Adaptive Self-Play for Software Engineering Agents
AlphaEvolve
ProjectGoogle DeepMind
AlphaEvolve
  1. Gemini models propose code changes
  2. Evaluators keep best scoring versions
  3. Finds better Gemini training kernels
  4. Lowers cost to train next Gemini models
  5. repeat
Kernels and compilersGoogle DeepMind · 2025
Darwin Gödel Machine
PaperarXiv
Darwin Gödel Machine
  1. Coding agent edits its own code
  2. Scored child agents enter the archive
  3. Archived agents make later self modifications
  4. repeat
Self-modifying agentsUBC, Vector Institute, Sakana AI · 2025
AlphaChip
ProjectGoogle DeepMind
AlphaChip
  1. RL agent lays out TPU blocks
  2. Better layouts make AI hardware cheaper
  3. Agent pre-trains on earlier chip generations
  4. repeat
Hardware and chipsGoogle DeepMind, Google Research · 2020
autoresearch
Mini-projectGitHub
autoresearch
  1. Coding agent edits LLM training script
  2. Runs a 5-minute training job
  3. Keeps change if validation improves
  4. Changes transferred to larger models
  5. repeat
Research automationAndrej Karpathy · 2026
ScholarEvolve
PaperarXiv
ScholarEvolve
  1. Research agent writes harness mutation blueprints
  2. Coding agent implements harness module changes
  3. Winning harness becomes the new champion
  4. New champion harness runs further generations
  5. repeat
Tools and harnessYang et al. (UC Santa Barbara, Microsoft) · 2026
Nemotron-4 340B
PaperarXiv
Nemotron-4 340B
  1. Aligned model generates training data
  2. Reward model judges and filters it
  3. Data trains next intermediate instruct model
  4. Next model generates better alignment data
  5. repeat
Training dataNVIDIA · 2024
Self-Rewarding LMs
PaperarXiv
Self-Rewarding LMs
  1. Model generates candidate responses to prompts
  2. Model scores them itself as a judge
  3. Scores become preference pairs for DPO
  4. Trained model becomes next generator and judge
  5. repeat
Self-reward and self-playMeta, NYU (Yuan et al.) · 2024
Claude Code
ProjectAnthropic
Claude Code
  1. Claude Code writes its own code changes
  2. Changes become next Claude Code versions
  3. Release becomes harness for next development round
  4. repeat
Self-modifying agentsAnthropic · 2025
KernelEvolve
ProjectarXiv
KernelEvolve
  1. LLM agents write and benchmark kernels
  2. Kernels speed up AI model training
  3. Session data post-trains smaller kernel models
  4. repeat
Kernels and compilersMeta · 2025
PostTrainBench
BenchmarkarXiv
PostTrainBench
  1. Coding agent writes fine-tuning code
  2. Agent post-trains a base LLM
  3. Trained model scored on target benchmark
  4. One turn of model training another
  5. repeat
AI R&D evalsELLIS Institute Tubingen, MPI-IS, University of Tubingen, Thoughtful Lab (Rank et al.) · 2026
Self-training learned optimizers
FoundationarXiv
Self-training learned optimizers
  1. Learned optimizers update other learned optimizers
  2. Training keeps best performing optimizer parameters
  3. Better optimizers train each other faster
  4. repeat
Architecture and optimizer searchMetz et al. (Google Research, Brain Team) · 2021
CoT monitoring
PaperarXiv
CoT monitoring
  1. LLM reads reasoning model thoughts
  2. LLM flags reward hacking behavior
  3. Signal feeds back into training reward
  4. Trains more aligned capable agents
  5. repeat
Interpretability and oversightOpenAI (Baker et al.) · 2025

Support for loop research

How to apply →

Have a loop worth building?

Tell us what the AI improves and how the improvement comes back. Applications for grants, fellowships and compute all start on the contact page.

Get in touch