https://rewire.it Tech consultancy, app building, creative engineering, and musings on philosophy and technology en Mon, 22 Jun 2026 10:39:53 GMT https://rewire.it/blog/fifteen-ways-to-fold-a-protein https://rewire.it/blog/fifteen-ways-to-fold-a-protein Thu, 18 Jun 2026 00:00:00 GMT AlphaFold3 reported roughly doubled accuracy on some interaction categories, then released its weights under a non-commercial license gated behind a Google form. Three open models have since matched it. The right choice is not the leaderboard winner; it is the one that fits your task and your license. protein-structure-prediction alphafold boltz foundation-models structural-biology drug-discovery machine-learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/antibody-specific-foundation-models https://rewire.it/blog/antibody-specific-foundation-models Wed, 17 Jun 2026 00:00:00 GMT Five of the ten leading antibody-specific models lack a clean permissive license for commercial use, and the hardest loop in an antibody still lands around 3 Angstroms across every structure predictor on the list. Those two numbers, the license and the loop, reorder the whole shortlist. antibody foundation-models protein-language-models structure-prediction antibody-design CDR-H3 OAS machine-learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/protein-design-foundation-models-in-2026 https://rewire.it/blog/protein-design-foundation-models-in-2026 Tue, 16 Jun 2026 00:00:00 GMT RFdiffusion2 scaffolded all 41 of 41 enzyme active sites on a benchmark where the prior best deep-learning method managed 16. That gap is the field in one number: diffusion made backbone generation almost routine on paper, and the slow, honest work is finding out how much survives contact with a pipette. Here are the 14 models, sorted by what they can do, what survives the wet lab, and whether you can legally ship them. protein-design foundation-models generative-models diffusion-models RFdiffusion ProteinMPNN bioinformatics protein-engineering tim@rewire.it (Tim Richardson) https://rewire.it/blog/protein-language-models-in-2026 https://rewire.it/blog/protein-language-models-in-2026 Tue, 16 Jun 2026 00:00:00 GMT ESM3 generated a fluorescent protein 58% identical to the nearest natural one, which the authors framed as 500 million years of evolution skipped. That headline is real. So is the fact that for most downstream tasks a well-tuned ESM-2 650M still wins on cost per useful prediction. This survey keeps two ledgers. protein-language-models foundation-models bioinformatics ESM machine-learning computational-biology tim@rewire.it (Tim Richardson) https://rewire.it/blog/genomic-foundation-models-in-2026 https://rewire.it/blog/genomic-foundation-models-in-2026 Wed, 10 Jun 2026 00:00:00 GMT The pitch for genomic foundation models is that one pretrained network now beats task-specific tools across the board, from regulatory annotation to clinical variant interpretation. genomics foundation-models clinical-genomics variant-effect-prediction benchmarking reproducibility molecular-pathology tim@rewire.it (Tim Richardson) https://rewire.it/blog/carbon-3b https://rewire.it/blog/carbon-3b Fri, 29 May 2026 00:00:00 GMT Carbon-3B matches Evo2-7B on sequence recovery, variant-effect prediction, and motif-perturbation discrimination while generating DNA over 150 times faster. genomics foundation-models dna-language-models tokenization transformers variant-effect-prediction tim@rewire.it (Tim Richardson) https://rewire.it/blog/mimic https://rewire.it/blog/mimic Fri, 29 May 2026 00:00:00 GMT A 1-billion-parameter model conditioned on RNA chemical-probing reactivity folds a held-out transcript to an F1 of 0.987 against an experimentally guided reference. foundation-models computational-biology multimodal splicing protein-design tim@rewire.it (Tim Richardson) https://rewire.it/blog/structure-without-alignment https://rewire.it/blog/structure-without-alignment Fri, 29 May 2026 00:00:00 GMT ESMFold predicts atomic-level structure from a single sequence. No multiple sequence alignment, no database search, no Evoformer churning over homologs. protein-language-models structure-prediction ESMFold ESM-2 metagenomics deep-learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/fewer-than-01-percent-of-neurons-predict-when-llms-hallucinate https://rewire.it/blog/fewer-than-01-percent-of-neurons-predict-when-llms-hallucinate Wed, 25 Feb 2026 00:00:00 GMT Fewer than one in a thousand neurons in a large language model can predict whether it's about to hallucinate -- and they encode something unexpected: not factual errors, but a tendency toward compliance over truth. mechanistic-interpretability hallucination llm-internals neurons alignment tim@rewire.it (Tim Richardson) https://rewire.it/blog/ai-safety-via-debate https://rewire.it/blog/ai-safety-via-debate Fri, 20 Feb 2026 00:00:00 GMT Reinforcement learning works when you can check the answer. A chess engine wins or loses. A code-generation model passes or fails the test suite. ai-safety reinforcement-learning scalable-oversight debate alignment tim@rewire.it (Tim Richardson) https://rewire.it/blog/inverse-graphics-as-rl-environments https://rewire.it/blog/inverse-graphics-as-rl-environments Fri, 20 Feb 2026 00:00:00 GMT Vision-language models can describe a scene in paragraph-length detail and still fail to tell you whether a red cube is in front of or behind a blue cylinder. reinforcement-learning vision-language-models inverse-graphics benchmarks spatial-reasoning tim@rewire.it (Tim Richardson) https://rewire.it/blog/socrates-was-a-terrible-prompt-engineer https://rewire.it/blog/socrates-was-a-terrible-prompt-engineer Wed, 11 Feb 2026 00:00:00 GMT Six ancient questioning techniques map onto the most effective LLM prompting strategies with uncomfortable precision, and what that reveals about these models is more interesting than the performance gains. prompt-engineering socratic-method llm-reasoning ai-philosophy chain-of-thought tim@rewire.it (Tim Richardson) https://rewire.it/blog/machine-readable-science https://rewire.it/blog/machine-readable-science Wed, 04 Feb 2026 00:00:00 GMT How do you verify scientific knowledge when there are 2.9 million papers on arXiv alone, with thousands more added every day? A new paper extracts nearly two million claims from 16,087 manuscripts and compares machine evaluation to human peer review, with 81% agreement. machine-learning scientific-publishing AI knowledge-graphs epistemology tim@rewire.it (Tim Richardson) https://rewire.it/blog/when-ai-writes-the-code-verification-becomes-the-job https://rewire.it/blog/when-ai-writes-the-code-verification-becomes-the-job Wed, 04 Feb 2026 00:00:00 GMT Over 80% of developers now use AI assistants for code generation, yet at least 62% of AI-generated code contains vulnerabilities. As AI writes code faster than humans can review it, the engineer's primary job shifts from writing code to verifying it through formal methods. formal-verification ai-code-generation software-security devops llm tim@rewire.it (Tim Richardson) https://rewire.it/blog/conglude-contrastive-drug-design https://rewire.it/blog/conglude-contrastive-drug-design Fri, 30 Jan 2026 16:00:00 GMT Structure-based and ligand-based drug design evolved as separate fields solving the same problem. ConGLUDe, a contrastive geometric learning model, unifies both approaches and outperforms specialist methods on realistic benchmarks without requiring pre-defined binding pockets. Drug Discovery Contrastive Learning Computational Biology Virtual Screening Protein-Ligand Interactions Graph Neural Networks Machine Learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/alphagenome-one-model-for-the-other-98-percent-of-your-dna https://rewire.it/blog/alphagenome-one-model-for-the-other-98-percent-of-your-dna Thu, 29 Jan 2026 16:00:00 GMT Google DeepMind's AlphaGenome reads 1 million base pairs of DNA and predicts thousands of regulatory functions at single-nucleotide resolution, beating 25 of 26 specialized models. deep-learning genomics alphagenome deepmind variant-prediction non-coding-dna tim@rewire.it (Tim Richardson) https://rewire.it/blog/alphagenome-variant-effect-prediction https://rewire.it/blog/alphagenome-variant-effect-prediction Thu, 29 Jan 2026 12:00:00 GMT AlphaGenome processes 1 million DNA base pairs to predict variant effects across 7,000+ genomic tracks in one second, outperforming specialized models on 25 of 26 VEP benchmarks. Genomics Deep Learning Variant Effect Prediction AlphaGenome Computational Biology tim@rewire.it (Tim Richardson) https://rewire.it/blog/alphagenome-gene-regulation-2d-embeddings-splicing-noncoding-dna https://rewire.it/blog/alphagenome-gene-regulation-2d-embeddings-splicing-noncoding-dna Thu, 29 Jan 2026 11:00:00 GMT A technical look at AlphaGenome's architecture, its 2D pairwise embeddings for splicing prediction, and what the model means for clinical variant interpretation. AlphaGenome Genomics Deep Learning Splicing Computational Biology Google DeepMind Variant Interpretation tim@rewire.it (Tim Richardson) https://rewire.it/blog/eden-28-billion-parameters-for-programming-biology https://rewire.it/blog/eden-28-billion-parameters-for-programming-biology Wed, 28 Jan 2026 14:00:00 GMT Basecamp Research's EDEN model trains on proprietary environmental metagenomics to design gene-insertion enzymes, antimicrobial peptides, and synthetic microbiomes -- all validated in the wet lab. Foundation Models Computational Biology Gene Therapy Metagenomics Drug Discovery EDEN Basecamp Research tim@rewire.it (Tim Richardson) https://rewire.it/blog/a-bioinformaticians-guide-to-choosing-genomic-foundation-models https://rewire.it/blog/a-bioinformaticians-guide-to-choosing-genomic-foundation-models Mon, 19 Jan 2026 10:30:00 GMT A practical guide to selecting genomic foundation models for bioinformatics tasks. Covers ESM-2, DNABERT-2, HyenaDNA, Nucleotide Transformer, scGPT, and Evo with specific recommendations for DNA sequence analysis, protein structure prediction, and single-cell analysis based on hardware requirements, inference speed, and task type. Foundation Models Genomics Bioinformatics Deep Learning Protein Language Models DNA Models ESM-2 DNABERT-2 HyenaDNA scGPT tim@rewire.it (Tim Richardson) https://rewire.it/blog/end-to-end-test-time-training-long-context-constant-latency https://rewire.it/blog/end-to-end-test-time-training-long-context-constant-latency Thu, 15 Jan 2026 11:00:00 GMT How TTT-E2E achieves constant inference latency regardless of context length by treating long context as a learning problem rather than an architecture problem. LLM Long Context Test-Time Training Machine Learning Transformers Inference Optimization tim@rewire.it (Tim Richardson) https://rewire.it/blog/engram-how-deepseek-added-second-brain-to-llm https://rewire.it/blog/engram-how-deepseek-added-second-brain-to-llm Tue, 13 Jan 2026 11:00:00 GMT A technical deep dive into DeepSeek's Engram architecture, which introduces conditional memory as a new axis of sparsity for large language models. Deep Learning LLM Architecture Memory Mixture of Experts DeepSeek Sparse Computation tim@rewire.it (Tim Richardson) https://rewire.it/blog/chemberta-when-language-models-learned-to-speak-chemistry https://rewire.it/blog/chemberta-when-language-models-learned-to-speak-chemistry Thu, 08 Jan 2026 11:00:00 GMT How the same transformer architecture powering GPT learned to predict molecular properties by treating chemistry as a language problem Machine Learning Chemistry Transformers Drug Discovery Molecular Property Prediction tim@rewire.it (Tim Richardson) https://rewire.it/blog/molbert-teaching-transformers-to-read-chemistry https://rewire.it/blog/molbert-teaching-transformers-to-read-chemistry Thu, 08 Jan 2026 10:00:00 GMT How researchers adapted BERT for molecular property prediction, turning SMILES strings into drug discovery insights Machine Learning Drug Discovery Transformers Chemistry SMILES BERT tim@rewire.it (Tim Richardson) https://rewire.it/blog/diffusion-surrogates-for-agent-based-biological-simulations https://rewire.it/blog/diffusion-surrogates-for-agent-based-biological-simulations Sat, 03 Jan 2026 12:00:00 GMT How generative diffusion models can serve as fast surrogates for expensive biological simulations, achieving 22x speedup while preserving the stochastic diversity that makes these models scientifically useful. Diffusion Models Agent-Based Modeling Computational Biology Cellular Potts Surrogates Machine Learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/ml-interpretability-precision-oncology https://rewire.it/blog/ml-interpretability-precision-oncology Sat, 03 Jan 2026 10:00:00 GMT Machine learning models now outperform FDA-approved biomarkers in predicting treatment response, but the best-performing models often resist explanation. Here's how precision oncology is navigating the trade-off between performance and interpretability. Machine Learning Precision Oncology Explainability Genomics Graph Neural Networks Clinical AI tim@rewire.it (Tim Richardson) https://rewire.it/blog/does-rl-actually-make-llms-smarter https://rewire.it/blog/does-rl-actually-make-llms-smarter Tue, 30 Dec 2025 13:00:00 GMT Recent research suggests RL training optimizes search efficiency over existing capabilities rather than expanding reasoning capacity. Here's what the pass@k evidence actually shows. Reinforcement Learning LLM Reasoning RLVR Machine Learning AI Research tim@rewire.it (Tim Richardson) https://rewire.it/blog/vl-jepa-why-predicting-embeddings-beats-generating-tokens https://rewire.it/blog/vl-jepa-why-predicting-embeddings-beats-generating-tokens Tue, 30 Dec 2025 12:00:00 GMT VL-JEPA achieves 50% parameter reduction and 2.85x faster decoding by predicting embeddings instead of generating tokens, offering a compelling alternative to autoregressive vision-language models. Vision Language JEPA Deep Learning Multimodal AI Self-Supervised Learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/project-silicon-gradient-descent-on-assembly-code https://rewire.it/blog/project-silicon-gradient-descent-on-assembly-code Tue, 30 Dec 2025 11:00:00 GMT A deep dive into Project Silicon's proposal to build differentiable CPU simulators, enabling gradient-based optimization of assembly code and opening a new frontier in neural algorithm synthesis. Machine Learning Systems Programming Program Synthesis Reinforcement Learning Differentiable Computing tim@rewire.it (Tim Richardson) https://rewire.it/blog/code-world-models-teaching-llms-to-simulate-execution https://rewire.it/blog/code-world-models-teaching-llms-to-simulate-execution Thu, 18 Dec 2025 21:00:00 GMT How Meta's Code World Model applies the "dreaming car" insight from robotics to software engineering, achieving 65.8% on SWE-bench Verified by training on execution traces rather than static code. Code Generation World Models LLM Agentic AI Software Engineering Meta AI tim@rewire.it (Tim Richardson) https://rewire.it/blog/biological-world-models-projects-you-should-build https://rewire.it/blog/biological-world-models-projects-you-should-build Thu, 18 Dec 2025 14:00:00 GMT Why computational biologists should stop building embeddings and start building simulators, with three tractable project ideas you can implement today using flow matching, Neural ODEs, and cell fate trajectory modeling. World Models Computational Biology Machine Learning Single Cell Neural ODE Flow Matching tim@rewire.it (Tim Richardson) https://rewire.it/blog/benchmarks-vs-rl-environments-why-the-distinction-matters https://rewire.it/blog/benchmarks-vs-rl-environments-why-the-distinction-matters Thu, 18 Dec 2025 12:00:00 GMT Understanding when you're working with an environment versus a benchmark changes how you design experiments, interpret results, and communicate findings. This guide covers the practical differences every RL practitioner should know. Reinforcement Learning Benchmarks Machine Learning Research Methodology tim@rewire.it (Tim Richardson) https://rewire.it/blog/why-1000-layer-networks-finally-work-for-reinforcement-learning https://rewire.it/blog/why-1000-layer-networks-finally-work-for-reinforcement-learning Thu, 18 Dec 2025 12:00:00 GMT Recent research shows 1024-layer networks achieve 2x to 50x improvements in goal-conditioned RL. Here's why extreme depth works now, and when you should consider it for your own agents. Reinforcement Learning Deep Learning Goal-Conditioned RL Network Architecture Self-Supervised Learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/from-diffusion-to-clinic-how-ai-is-designing-antibodies-from-scratch https://rewire.it/blog/from-diffusion-to-clinic-how-ai-is-designing-antibodies-from-scratch Tue, 09 Dec 2025 11:00:00 GMT AI-designed antibodies have achieved atomic-level structural accuracy and therapeutic-relevant binding affinities, compressing discovery timelines from years to weeks. antibody-design machine-learning drug-discovery diffusion-models AlphaFold3 tim@rewire.it (Tim Richardson) https://rewire.it/blog/the-stability-plasticity-dilemma-how-memory-architectures-are-solving-continual-learning https://rewire.it/blog/the-stability-plasticity-dilemma-how-memory-architectures-are-solving-continual-learning Tue, 09 Dec 2025 10:00:00 GMT A comparative analysis of approaches to catastrophic forgetting in language models, from parameter regularization to sparse memory architectures that reduce forgetting from 89% to just 11%. continual-learning catastrophic-forgetting memory-layers machine-learning llm tim@rewire.it (Tim Richardson) https://rewire.it/blog/disco-rl-when-algorithms-learn-to-design-algorithms https://rewire.it/blog/disco-rl-when-algorithms-learn-to-design-algorithms Mon, 08 Dec 2025 00:00:00 GMT DeepMind's DiscoRL discovers reinforcement learning algorithms that outperform hand-designed methods like PPO and DQN. By treating algorithm design as a meta-learning problem, it found alternatives to value functions and bootstrapping through optimization alone. Reinforcement Learning Meta-Learning DeepMind Algorithm Discovery AI Research Machine Learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/do-llms-construct-world-models-a-cognitive-science-investigation https://rewire.it/blog/do-llms-construct-world-models-a-cognitive-science-investigation Mon, 08 Dec 2025 00:00:00 GMT Are large language models merely stochastic parrots, or do they develop genuine internal representations of the world? This investigation examines evidence from Othello-GPT, spatial encoding in LLMs, and the symbol grounding problem to explore what cognitive science reveals about AI understanding. Cognitive Science LLMs World Models Philosophy of Mind AI Research GPT-4 Symbol Grounding Machine Learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/mixture-of-experts-the-efficiency-trick-behind-modern-ai https://rewire.it/blog/mixture-of-experts-the-efficiency-trick-behind-modern-ai Mon, 08 Dec 2025 00:00:00 GMT Mixtral uses 46.7B parameters but only activates 13B per token. This architectural trick called Mixture of Experts powers Gemini 1.5, DeepSeek V3, and more. Learn how MoE works, its hidden costs, and when to use it. Machine Learning MoE Efficiency LLM Architecture Mixtral DeepSeek Neural Networks tim@rewire.it (Tim Richardson) https://rewire.it/blog/tensor-logic-one-equation-to-rule-them-all https://rewire.it/blog/tensor-logic-one-equation-to-rule-them-all Mon, 08 Dec 2025 00:00:00 GMT Pedro Domingos proposes that neural networks and symbolic AI are the same mathematical operation - a logical rule can be equivalently written as a tensor equation in Einstein summation notation. If true, we've been building separate tools for problems that share identical structure. Tensor Logic Neuro-Symbolic AI Einstein Notation Transformers Deep Learning Machine Learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/when-machines-design-their-own-learning-algorithms https://rewire.it/blog/when-machines-design-their-own-learning-algorithms Mon, 08 Dec 2025 00:00:00 GMT A machine trained on simple grid worlds beat every hand-designed RL algorithm on Atari. DeepMind's DiscoRL discovers algorithms through meta-learning that outperform DQN, PPO, and A3C - methods humans spent decades developing. Reinforcement Learning Meta-Learning Algorithm Discovery Deep Learning AI Research Machine Learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/why-your-llm-only-uses-10-20-percent-of-its-context-window-and-how-titans-fixes-it https://rewire.it/blog/why-your-llm-only-uses-10-20-percent-of-its-context-window-and-how-titans-fixes-it Mon, 08 Dec 2025 00:00:00 GMT GPT-4's 128K context window? It only uses about 10% effectively. Google's TITANS architecture introduces test-time memory learning that outperforms GPT-4 on long-context tasks with 70x fewer parameters. AI Machine Learning Transformers Memory Architectures Long Context TITANS MIRAS Neural Networks tim@rewire.it (Tim Richardson) https://rewire.it/blog/your-brain-on-rhythm-what-39-cultures-taught-scientists-about-universal-music https://rewire.it/blog/your-brain-on-rhythm-what-39-cultures-taught-scientists-about-universal-music Thu, 04 Dec 2025 00:00:00 GMT A landmark study across five continents reveals that all humans share a preference for mathematically simple rhythms, but the specific patterns we favor are shaped by the music we absorb growing up. Music Cognition Cross-Cultural Research Neuroscience Rhythm Perception tim@rewire.it (Tim Richardson) https://rewire.it/blog/biologys-secret-weapon-physics-based-benchmarks-for-training-rl-agents https://rewire.it/blog/biologys-secret-weapon-physics-based-benchmarks-for-training-rl-agents Wed, 03 Dec 2025 00:00:00 GMT Why biological systems offer the ideal training ground for reinforcement learning: automated verification through physics, not human judgment. From protein design with AlphaFold to RNA folding with ViennaRNA, biology provides the verifiable inverse problems that RL needs at scale. Reinforcement Learning Biological Benchmarks Protein Design RNA Design Computational Biology Machine Learning AlphaFold Drug Discovery tim@rewire.it (Tim Richardson) https://rewire.it/blog/the-dark-matter-of-biology-what-machine-learning-reveals-about-the-invisible-proteome https://rewire.it/blog/the-dark-matter-of-biology-what-machine-learning-reveals-about-the-invisible-proteome Tue, 02 Dec 2025 00:00:00 GMT Machine learning is illuminating biology's hidden half: intrinsically disordered proteins and RNA structures that traditional methods could never capture. But can we trust what we're seeing? Machine Learning Computational Biology Protein Structure RNA AlphaFold Intrinsic Disorder tim@rewire.it (Tim Richardson) https://rewire.it/blog/what-are-world-models-ai-path-to-understanding-reality https://rewire.it/blog/what-are-world-models-ai-path-to-understanding-reality Thu, 13 Nov 2025 00:00:00 GMT World models enable AI agents to imagine futures and plan actions, achieving 10-100x better sample efficiency than traditional reinforcement learning. From DreamerV3 collecting diamonds in Minecraft to foundation models like Sora and Genie, world models represent AI's shift from pattern matching to simulating reality itself. AI Machine Learning Reinforcement Learning World Models Robotics Autonomous Vehicles tim@rewire.it (Tim Richardson) https://rewire.it/blog/have-we-reached-human-level-agi-what-the-evidence-actually-shows https://rewire.it/blog/have-we-reached-human-level-agi-what-the-evidence-actually-shows Wed, 12 Nov 2025 13:56:00 GMT Smart people disagree dramatically about whether AGI is five years away or fifty. The evidence reveals why this definitional crisis matters more than the technology itself. AGI artificial-intelligence evidence-based expert-analysis AI-debate tim@rewire.it (Tim Richardson) https://rewire.it/blog/ai-designed-two-new-antibiotics-from-scratch-heres-why-that-changes-everything https://rewire.it/blog/ai-designed-two-new-antibiotics-from-scratch-heres-why-that-changes-everything Wed, 12 Nov 2025 12:57:00 GMT MIT researchers used generative AI to create novel antibiotics from scratch, not finding them, but designing them. The breakthrough matters less for what it creates than for what it proves. AI antibiotics drug-discovery synthetic-biology biotechnology generative-AI antimicrobial-resistance tim@rewire.it (Tim Richardson) https://rewire.it/blog/how-to-build-a-dspy-application-from-prompt-whack-a-mole-to-systematic-optimization https://rewire.it/blog/how-to-build-a-dspy-application-from-prompt-whack-a-mole-to-systematic-optimization Wed, 12 Nov 2025 10:00:00 GMT Learn how DSPy transforms brittle prompt engineering into systematic, testable code with automatic optimization dspy ai-engineering llm-programming prompt-optimization tutorial tim@rewire.it (Tim Richardson) https://rewire.it/blog/nested-learning-how-your-neural-network-already-learns-at-multiple-timescales https://rewire.it/blog/nested-learning-how-your-neural-network-already-learns-at-multiple-timescales Mon, 10 Nov 2025 18:00:00 GMT Nested Learning: The Illusion of Deep Learning Architectures - A comprehensive guide to the arXiv paper revealing how neural networks learn at multiple timescales through hierarchical optimization. deep-learning neural-networks optimization memory-consolidation language-models transformers continual-learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/the-amnesia-problem-why-neural-networks-cant-learn-like-humans https://rewire.it/blog/the-amnesia-problem-why-neural-networks-cant-learn-like-humans Mon, 10 Nov 2025 12:00:00 GMT Neural networks catastrophically forget previous knowledge when learning new tasks—not due to capacity limits but fundamental constraints in distributed learning systems. machine-learning neural-networks continual-learning theory tim@rewire.it (Tim Richardson) https://rewire.it/blog/alphaevolve-breaking-56-years-of-mathematical-stagnation https://rewire.it/blog/alphaevolve-breaking-56-years-of-mathematical-stagnation Thu, 06 Nov 2025 18:00:00 GMT Google DeepMind's AlphaEvolve broke a 56-year-old matrix multiplication record and matched or beat human solutions on 95% of 67 mathematical problems. alphaevolve machine-learning algorithm-discovery mathematics deepmind tim@rewire.it (Tim Richardson) https://rewire.it/blog/large-language-models-and-emergence-a-complex-systems-perspective https://rewire.it/blog/large-language-models-and-emergence-a-complex-systems-perspective Thu, 06 Nov 2025 14:00:00 GMT Are LLMs truly exhibiting emergent capabilities, or are we mistaking measurement artifacts for genuine phase transitions? AI LLMs emergence complexity-science machine-learning tim@rewire.it (Tim Richardson) https://rewire.it/blog/why-gene-discovery-methods-find-different-genes https://rewire.it/blog/why-gene-discovery-methods-find-different-genes Thu, 06 Nov 2025 10:00:00 GMT How gene length, trait specificity, and luck systematically distort which genes we think are important for disease, and what researchers should do about it genomics GWAS bioinformatics drug-discovery statistical-genetics tim@rewire.it (Tim Richardson) https://rewire.it/blog/kosmos-12-hour-ai-research-session https://rewire.it/blog/kosmos-12-hour-ai-research-session Thu, 06 Nov 2025 00:00:00 GMT An AI research system maintains coherent reasoning across 200+ agent steps over 12 hours, generating 42,000 lines of code while reviewing 1,500 papers. Kosmos achieves 79.4% accuracy through structured world models and parallel agents, but verification remains humanity's bottleneck. ai-research autonomous-agents machine-learning scientific-discovery ai-systems tim@rewire.it (Tim Richardson) https://rewire.it/blog/the-complexity-cliff-why-reasoning-models-work-until-they-dont https://rewire.it/blog/the-complexity-cliff-why-reasoning-models-work-until-they-dont Wed, 05 Nov 2025 21:59:08 GMT New research reveals a disturbing truth: reasoning models maintain high performance until they hit a complexity threshold, then collapse entirely. They don't degrade gracefully - they fall off a cliff. AI Machine Learning Research LLMs Reasoning tim@rewire.it (Tim Richardson) https://rewire.it/blog/why-foundation-models-in-pathology-are-failing-and-what-comes-next https://rewire.it/blog/why-foundation-models-in-pathology-are-failing-and-what-comes-next Wed, 29 Oct 2025 08:01:00 GMT Why does it feel like our tools weren't designed by pathologists? Billions poured into AI models that compress whole slide images into tiny vectors, ignoring how pathologists actually examine tissue. The evidence reveals why scaling won't fix this disconnect. pathology foundation-models medical-AI deep-learning multiple-instance-learning clinical-validation digital-pathology machine-learning computer-vision tim@rewire.it (Tim Richardson) https://rewire.it/blog/building-metacognitive-ai-agents-complete-guide https://rewire.it/blog/building-metacognitive-ai-agents-complete-guide Tue, 05 Nov 2024 16:50:51 GMT The definitive guide to building AI agents that monitor their own thinking through dual-loop architecture, Reflexion patterns, and production-ready LangGraph implementations. AI Agent Metacognition LangGraph Safety Python Production Tutorial tim@rewire.it (Tim Richardson) https://rewire.it/blog/ai-discovers-bacterial-survival-strategy-cf-pici https://rewire.it/blog/ai-discovers-bacterial-survival-strategy-cf-pici Tue, 05 Nov 2024 09:29:56 GMT An AI system at Google DeepMind discovered how bacteria share genes across species barriers using 7 days of computational reasoning. When tested, it matched unpublished experimental observations exactly. AI Biology Research Machine Learning Bacteria Gene Transfer DeepMind tim@rewire.it (Tim Richardson) https://rewire.it/blog/foundation-models-are-rewriting-the-rules-of-biology https://rewire.it/blog/foundation-models-are-rewriting-the-rules-of-biology Wed, 30 Oct 2024 11:35:02 GMT Foundation models trained on biological data are transforming protein structure prediction, genomics, drug discovery, and pathology. Learn how machine learning benchmarks in 2024 are revealing biology's dark matter through RNA analysis and metagenomic discovery. AI Machine Learning Biology Bioinformatics Foundation Models AlphaFold Drug Discovery Dark Matter Biology RNA Benchmarks Metagenomics tim@rewire.it (Tim Richardson) https://rewire.it/blog/most-ai-agents-arent-actually-agents-autonomy-framework https://rewire.it/blog/most-ai-agents-arent-actually-agents-autonomy-framework Mon, 28 Oct 2024 20:29:09 GMT How the 6-level autonomy framework helps technical managers distinguish genuine autonomous systems from marketing claims and avoid expensive pilots that were never going to work. AI Agents Autonomy Technical Evaluation Enterprise AI Data Agents tim@rewire.it (Tim Richardson) https://rewire.it/blog/claude-code-agents-skills-slash-commands https://rewire.it/blog/claude-code-agents-skills-slash-commands Sat, 26 Oct 2024 20:52:18 GMT Skills vs Slash Commands: understand the key differences, when to use each, and how they work together in Claude Code. Learn which abstraction fits your workflow with practical examples and decision frameworks. claude-code ai-agents anthropic developer-tools llm agentic-systems tim@rewire.it (Tim Richardson) https://rewire.it/blog/claudes-powerful-toolkit-for-life-sciences https://rewire.it/blog/claudes-powerful-toolkit-for-life-sciences Fri, 25 Oct 2024 21:40:51 GMT Why does it feel like our tools weren't designed by pathologists or researchers? Anthropic's Claude for Life Sciences attempts to bridge this gap by embedding AI directly into existing scientific workflows through Connectors and Agent Skills, building an operating system for R&D that actually respects how scientists work. ai life-sciences anthropic bioinformatics claude drug-discovery tim@rewire.it (Tim Richardson) https://rewire.it/blog/using-embedding-models-to-predict-sentence-complexity https://rewire.it/blog/using-embedding-models-to-predict-sentence-complexity Thu, 24 Oct 2024 21:41:07 GMT Traditional readability formulas miss the mark. Modern embedding models can capture semantic nuance and syntactic structure, but do they actually predict complexity better? nlp embeddings machine-learning readability bert transformers tim@rewire.it (Tim Richardson) https://rewire.it/blog/from-structure-to-function-leveraging-alphafolds-evoformer-embeddings-for-downstream-ai https://rewire.it/blog/from-structure-to-function-leveraging-alphafolds-evoformer-embeddings-for-downstream-ai Thu, 24 Oct 2024 18:27:41 GMT AlphaFold solved protein folding, but its hidden embeddings may be even more valuable, powering everything from drug design to disease prediction. alphafold protein-embeddings machine-learning computational-biology evoformer structural-biology tim@rewire.it (Tim Richardson) https://rewire.it/blog/the-embedding-dilemma-why-your-rag-fails-and-how-to-think-in-chunks https://rewire.it/blog/the-embedding-dilemma-why-your-rag-fails-and-how-to-think-in-chunks Thu, 24 Oct 2024 09:56:28 GMT Discover why monolithic embeddings fail for RAG systems and learn how chunking strategies can transform your retrieval performance. AI RAG Embeddings Vector Databases Machine Learning NLP tim@rewire.it (Tim Richardson) https://rewire.it/blog/beyond-the-page-the-rise-of-graph-structured-knowledge-in-language-models https://rewire.it/blog/beyond-the-page-the-rise-of-graph-structured-knowledge-in-language-models Tue, 15 Oct 2024 12:36:43 GMT How graph-based representations are transforming language models beyond sequential text to unlock deeper reasoning capabilities AI Machine Learning NLP Knowledge Graphs Graph Neural Networks tim@rewire.it (Tim Richardson) https://rewire.it/blog/dynamic-tool-allocation-for-ai-agents-the-rats-pattern https://rewire.it/blog/dynamic-tool-allocation-for-ai-agents-the-rats-pattern Wed, 04 Sep 2024 15:55:44 GMT Learn the RATS pattern: Retrieval-Augmented Tool Selection for building scalable AI agents that dynamically select the right tools for each task. AI Agents RAG RATS Tool Selection Architecture Google ADK tim@rewire.it (Tim Richardson) https://rewire.it/blog/beyond-the-subword-morphologically-aware-tokenization https://rewire.it/blog/beyond-the-subword-morphologically-aware-tokenization Tue, 03 Sep 2024 12:47:30 GMT Language models can write poetry but struggle with basic spelling. Discover why current tokenization breaks language, and how morphology-aware approaches fix it. nlp tokenization morphology linguistics ai language-models tim@rewire.it (Tim Richardson) https://rewire.it/blog/teaching-ai-structural-design https://rewire.it/blog/teaching-ai-structural-design Mon, 08 Jul 2024 11:54:48 GMT Exploring how Reinforcement Learning (RL) combined with Physics-Informed Machine Learning (PIML) can teach AI to design structurally sound and resilient buildings by learning from simulated physical environments. AI Reinforcement Learning Architecture Structural Engineering Physics Simulation PIDRL PIML tim@rewire.it (Tim Richardson) https://rewire.it/blog/geometry-of-meaning https://rewire.it/blog/geometry-of-meaning Tue, 02 Apr 2024 11:29:49 GMT Words aren't just symbols anymore, they're coordinates in high-dimensional space. Explore how vector embeddings revolutionized AI's understanding of language. AI Language Philosophy Vectors Cognitive Science tim@rewire.it (Tim Richardson)