Dogukan Tuna

Open-source tiny research lab focusing on generalization and continual learning continuaLM Lab — early stage, heavy research

Multi-agent system infrastructure at Manuel AI — HVAC, energy, marine

AI is the ultimate compressor of the world’s hardest problems— a superpower for scientific discovery with universal impact. The limit is no longer imagination. It is compute— and the absence of true algorithmic breakthroughs, scalable agents, and automated AI R&D. Toward generalization. Deployed Super-Learning Efficiency. Mechanistic Interpretability. March 2026.

Starting /now

✨

Starting /now — logging what I'm working on, building and thinking about

Mar 20, 2026

View all →

Mem-RLM — Memory-Augmented Inference for Recursive Language Models

An open-source memory layer for Recursive Language Models that records execution trajectories, extracts reusable strategies, and injects them into future runs. Models stop starting cold and actually learn which approaches work for which problem types — 26% accuracy improvement on weaker models, fully stateful inference.

Feb 23, 2026·6 min read·LLM inference, memory augmentation, recursive language models, reinforcement learning, AI agents, open source, test-time compute, strategy learning

A self-improving skill catalog for AI agents

An open-source skill catalog that agents use, extend, and improve themselves. 19 skills covering the full LLM lifecycle, autonomous research, GPU/TPU/QPU programming, and scientific computing — built by agents, for agents.

Mar 11, 2026·6 min read·GPU programming, TPU, QPU, quantum computing, HPC, CUDA, agent skills, open source, accelerated computing, LLM training, reinforcement learning, autonomous research

Claude Code-Time Skill Acquisition with Agent Teams

A team of agents researched, synthesized, and integrated a production-grade React Native skill into a shared knowledge base in under 15 minutes — just through coordination at Claude Code-time.

Feb 7, 2026·24 min read·AI agents, Claude Code, multi-agent systems, skill acquisition, agent coordination, knowledge base, React Native

On Compression, Computation and the Space Between

Kolmogorov complexity, neural networks as program search and Wolfram's ruliology seem to be looking at the same thing from different rooms.

Feb 1, 2026·12 min read·Kolmogorov complexity, information theory, computation theory, neural networks, Wolfram, compression, mathematics

Defeating Nondeterminism in LLM Inference: Reproducing Batch-Invariant Ops (RMSNorm & Tiled Matrix Multiplication) in JAX

This learning log is my beginning of a series exploring various kernel-related topics. As a starting point, I will reproduce the implementation of batch-invariant NN operations in JAX, drawing from Thinking Machines Lab's seminal collaborative work, \"Defeating Nondeterminism in LLM Inference.\"

Nov 25, 2025·26 min read·JAX, GPU kernels, RMSNorm, matrix multiplication, LLM inference, nondeterminism, batch invariance, deep learning

Streaming deepagents and task delegation with real-time output

This post demonstrates how to implement streaming capabilities on top of DeepAgents' package with multi-agent setup, with practical code examples and architectural patterns you can apply to your own projects.

Oct 20, 2025·9 min read·LLM agents, streaming, deep agents, multi-agent systems, task delegation, Python, AI engineering

Energetics of Allosteric Communication in Ubiquitin Revealed by Hybrid MCTS-Langevin Simulations

Exploring protein conformational landscapes and identifying potential allosteric communication pathways remain significant challenges in computational biophysics. This study presents a hybrid computational approach combining Monte Carlo Tree Search (MCTS) with Langevin Dynamics (LD) simulations using the OpenMM toolkit to enhance conformational sampling.

May 6, 2025·14 min read·computational biophysics, protein dynamics, Monte Carlo Tree Search, Langevin dynamics, OpenMM, molecular simulation, ubiquitin, allostery

Now Contact