{
  "total": 39,
  "reads": [
    {
      "id": "2026/08/prompt-injection-as-role-confusion",
      "slug": "2026/08/prompt-injection-as-role-confusion",
      "title": "Prompt Injection as Role Confusion",
      "url": "https://role-confusion.github.io/",
      "externalUrl": "https://role-confusion.github.io/",
      "date": "2026-08-16T00:00:00.000Z",
      "tags": [
        "ai-safety",
        "llm"
      ],
      "notes_markdown": "Shows that LLMs identify system vs. user roles by writing style rather than special delimit tags, and introduces **CoT Forgery** where models mistake injected instructions for their own reasoning trace.\n\n- Suggests that prompt injection defense requires semantic role separation rather than delimiter engineering.\n- Explains why chain-of-thought models remain vulnerable to deceptive thinking patterns.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/08/ai-detection-tools-cannot-prove-ai-text",
      "slug": "2026/08/ai-detection-tools-cannot-prove-ai-text",
      "title": "AI Detection Tools Cannot Prove That Text is AI-Generated",
      "url": "https://www.seangoedecke.com/ai-detection/",
      "externalUrl": "https://www.seangoedecke.com/ai-detection/",
      "date": "2026-08-14T00:00:00.000Z",
      "tags": [
        "ai-safety",
        "evals",
        "llm"
      ],
      "notes_markdown": "A critical breakdown of why statistical AI text classifiers suffer from high false-positive rates on formal, non-native, and technical English.\n\n- Explains why perplexity and burstiness metrics fail in real-world adversarial settings.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/08/beej-guide-to-network-programming",
      "slug": "2026/08/beej-guide-to-network-programming",
      "title": "Beej’s Guide to Network Programming",
      "url": "https://beej.us/guide/bgnet/html/split/",
      "externalUrl": "https://beej.us/guide/bgnet/html/split/",
      "date": "2026-08-14T00:00:00.000Z",
      "tags": [
        "systems",
        "software-engineering"
      ],
      "notes_markdown": "The canonical, timeless reference for POSIX socket programming, IP routing, client/server architectures, and low-level networking in C.\n\n- Essential fundamentals for distributed systems and custom network transports.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/08/how-ai-text-watermarking-works",
      "slug": "2026/08/how-ai-text-watermarking-works",
      "title": "How AI Text Watermarking Works: A Visual Guide",
      "url": "https://declaude.org/watermarking/",
      "externalUrl": "https://declaude.org/watermarking/",
      "date": "2026-08-14T00:00:00.000Z",
      "tags": [
        "ai-safety",
        "llm",
        "ml"
      ],
      "notes_markdown": "An intuitive visual guide to how statistical marks are embedded inside token sampling distributions without degrading generation quality.\n\n- Explains red/green token partitioning based on pseudo-random keys.\n- Shows what editing and paraphrasing transformations can erase the signal.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/08/nature-scalable-watermarking-llms",
      "slug": "2026/08/nature-scalable-watermarking-llms",
      "title": "Scalable Watermarking for Identifying Large Language Model Outputs",
      "url": "https://www.nature.com/articles/s41586-024-08025-4",
      "externalUrl": "https://www.nature.com/articles/s41586-024-08025-4",
      "date": "2026-08-14T00:00:00.000Z",
      "tags": [
        "ai-safety",
        "research",
        "llm"
      ],
      "notes_markdown": "DeepMind's foundational *Nature* paper on SynthID distortion-free cryptographic watermarking for generative models.\n\n- Proves statistical bounds for detection without altering token probability distributions.\n- Discusses production deployments across Gemini APIs.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/08/structured-decoding-with-xgrammar",
      "slug": "2026/08/structured-decoding-with-xgrammar",
      "title": "Structured Decoding with XGrammar",
      "url": "https://aramesh10.github.io/xgrammar/index.html",
      "externalUrl": "https://aramesh10.github.io/xgrammar/index.html",
      "date": "2026-08-14T00:00:00.000Z",
      "tags": [
        "llm",
        "systems",
        "developer-tools"
      ],
      "notes_markdown": "Deep dive into XGrammar's grammar-guided constrained decoding algorithms.\n\n- Efficiently pre-computes grammar masks across Context-Free Grammars (CFGs) and JSON schemas.\n- Achieves near-zero latency penalty during token generation.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/08/llm-inference-economics-from-first-principles",
      "slug": "2026/08/llm-inference-economics-from-first-principles",
      "title": "LLM Inference Economics from First Principles",
      "url": "https://www.tensoreconomics.com/p/llm-inference-economics-from-first",
      "externalUrl": "https://www.tensoreconomics.com/p/llm-inference-economics-from-first",
      "date": "2026-08-12T00:00:00.000Z",
      "tags": [
        "systems",
        "llm",
        "software-engineering"
      ],
      "notes_markdown": "Rigorous economic and hardware analysis of serving LLMs at scale.\n\n- Models memory bandwidth limits vs compute-bound FLOPs across prefill and decode stages.\n- Breaks down token cost structures and batching efficiency.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/08/lattice-static-retriever",
      "slug": "2026/08/lattice-static-retriever",
      "title": "Lattice: An 8 MB Static Retriever That Embeds Wikipedia in 7 Minutes",
      "url": "https://huggingface.co/blog/erikkaum/lattice-blog",
      "externalUrl": "https://huggingface.co/blog/erikkaum/lattice-blog",
      "date": "2026-08-11T00:00:00.000Z",
      "tags": [
        "rag",
        "systems",
        "ml"
      ],
      "notes_markdown": "An ultra-compact static embedding model requiring only 8 MB RAM.\n\n- Capable of indexing massive corpora on constrained edge hardware in minutes.\n- Demonstrates extreme quantization and lexical-dense hybridization.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/08/let-us-talk-about-encrypted-reasoning",
      "slug": "2026/08/let-us-talk-about-encrypted-reasoning",
      "title": "Let’s Talk About Encrypted Reasoning",
      "url": "https://blog.cryptographyengineering.com/2026/05/29/fooling-around-with-encrypted-reasoning-blobs/",
      "externalUrl": "https://blog.cryptographyengineering.com/2026/05/29/fooling-around-with-encrypted-reasoning-blobs/",
      "date": "2026-08-11T00:00:00.000Z",
      "tags": [
        "ai-safety",
        "llm",
        "research"
      ],
      "notes_markdown": "Matthew Green's analysis on encrypting chain-of-thought reasoning blobs returned by proprietary model APIs.\n\n- Explores how cryptographic encapsulation protects proprietary distillation while maintaining verifiable responses.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/08/networking-from-first-principles",
      "slug": "2026/08/networking-from-first-principles",
      "title": "Networking and the Internet, from First Principles",
      "url": "https://fazamhd.com/mental-models/networking/",
      "externalUrl": "https://fazamhd.com/mental-models/networking/",
      "date": "2026-08-03T00:00:00.000Z",
      "tags": [
        "systems",
        "software-engineering"
      ],
      "notes_markdown": "Clear mental models breaking down how packets traverse routers, switches, DNS lookups, TCP handshakes, and NAT tables.\n\n- Great refresher for backend and distributed systems engineering.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/07/autoresearch-and-metaharness",
      "slug": "2026/07/autoresearch-and-metaharness",
      "title": "Aman’s AI Journal • Primers • Autoresearch and Metaharness",
      "url": "https://aman.ai/primers/ai/autoresearch-and-metaharness/",
      "externalUrl": "https://aman.ai/primers/ai/autoresearch-and-metaharness/",
      "date": "2026-07-27T00:00:00.000Z",
      "tags": [
        "ai-agents",
        "research",
        "evals"
      ],
      "notes_markdown": "Detailed taxonomy of autonomous scientific research agents and the meta-harnesses required to validate hypotheses, run experiments, and synthesize literature.\n\n- Bridges theoretical agent architectures with practical laboratory workflows.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/07/agentic-world-models",
      "slug": "2026/07/agentic-world-models",
      "title": "Agentic World Models",
      "url": "https://cameronrwolfe.substack.com/p/agentic-world-models",
      "externalUrl": "https://cameronrwolfe.substack.com/p/agentic-world-models",
      "date": "2026-07-20T00:00:00.000Z",
      "tags": [
        "ai-agents",
        "rl",
        "ml"
      ],
      "notes_markdown": "Explores how generative environment simulators allow agents to simulate actions and consequences in latent space before committing real-world actions.\n\n- Essential reading for model-based RL and forward simulation in software engineering agents.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/07/is-frontier-asynchronous-rl-solved",
      "slug": "2026/07/is-frontier-asynchronous-rl-solved",
      "title": "Is Frontier Asynchronous RL Solved?",
      "url": "https://luk-huang.github.io/personal-website/blog/is-frontier-asynchronous-rl-solved.html",
      "externalUrl": "https://luk-huang.github.io/personal-website/blog/is-frontier-asynchronous-rl-solved.html",
      "date": "2026-07-15T00:00:00.000Z",
      "tags": [
        "rl",
        "systems",
        "research"
      ],
      "notes_markdown": "Luke Huang on the challenges of asynchronous, distributed policy gradient methods at frontier model scale.\n\n- Evaluates staleness penalties, off-policy correction heuristics, and gradient synchronization tradeoffs.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/07/continual-learning-with-rl-for-llms",
      "slug": "2026/07/continual-learning-with-rl-for-llms",
      "title": "Continual Learning with RL for LLMs",
      "url": "https://cameronrwolfe.substack.com/p/rl-continual-learning",
      "externalUrl": "https://cameronrwolfe.substack.com/p/rl-continual-learning",
      "date": "2026-07-08T00:00:00.000Z",
      "tags": [
        "rl",
        "fine-tuning",
        "ml"
      ],
      "notes_markdown": "Cameron Wolfe's analysis on applying reinforcement learning for ongoing post-training adaptation without catastrophic forgetting.\n\n- Explores replay buffers, KL regularizers, and distribution shift mitigation.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/07/decoding-speculative-decoding-first-principles",
      "slug": "2026/07/decoding-speculative-decoding-first-principles",
      "title": "Decoding Speculative Decoding from First Principles",
      "url": "https://jwlabs.vercel.app/post/speculative-decoding-first-principles",
      "externalUrl": "https://jwlabs.vercel.app/post/speculative-decoding-first-principles",
      "date": "2026-07-08T00:00:00.000Z",
      "tags": [
        "llm",
        "systems",
        "ml"
      ],
      "notes_markdown": "Step-by-step mathematical derivation and algorithmic mechanics of speculative decoding.\n\n- Covers draft model generation, target verification trees, and modified rejection sampling proofs.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/07/harness-engineering-for-self-improvement",
      "slug": "2026/07/harness-engineering-for-self-improvement",
      "title": "Harness Engineering for Self-Improvement",
      "url": "https://lilianweng.github.io/posts/2026-07-04-harness/",
      "externalUrl": "https://lilianweng.github.io/posts/2026-07-04-harness/",
      "date": "2026-07-07T00:00:00.000Z",
      "tags": [
        "ai-agents",
        "evals",
        "software-engineering"
      ],
      "notes_markdown": "Lilian Weng's comprehensive guide on building robust evaluation harnesses that allow AI agents to iteratively improve their own performance.\n\n- Focuses on environment feedback loops, automated failure analysis, and benchmark integrity.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/07/tinyrouter-llm-router-evolution",
      "slug": "2026/07/tinyrouter-llm-router-evolution",
      "title": "TinyRouter: A 10K-Parameter LLM Router Trained by Evolution",
      "url": "https://github.com/harrrshall/tinyrouter/",
      "externalUrl": "https://github.com/harrrshall/tinyrouter/",
      "date": "2026-07-06T00:00:00.000Z",
      "tags": [
        "developer-tools",
        "ml",
        "systems"
      ],
      "notes_markdown": "Harshal's ultra-compact model router trained with sep-CMA-ES.\n\n- Dynamically selects open-source models (DeepSeek, GLM, Kimi) based on query complexity and expected compute cost.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/07/from-harness-lock-in-to-portable-context-layer",
      "slug": "2026/07/from-harness-lock-in-to-portable-context-layer",
      "title": "From Harness Lock-In to Portable Context Layer",
      "url": "https://www.decodingai.com/p/the-context-layer",
      "externalUrl": "https://www.decodingai.com/p/the-context-layer",
      "date": "2026-07-05T00:00:00.000Z",
      "tags": [
        "ai-agents",
        "software-engineering",
        "developer-tools"
      ],
      "notes_markdown": "Architectural arguments for standardizing context storage and state machines independently of vendor-specific agent frameworks.\n\n- Emphasizes portable memory representations and interoperable agent tool interfaces.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/07/mastering-agentic-techniques-ai-agent-rl",
      "slug": "2026/07/mastering-agentic-techniques-ai-agent-rl",
      "title": "Mastering Agentic Techniques: AI Agent Reinforcement Learning",
      "url": "https://developer.nvidia.com/blog/mastering-agentic-techniques-ai-agent-reinforcement-learning/",
      "externalUrl": "https://developer.nvidia.com/blog/mastering-agentic-techniques-ai-agent-reinforcement-learning/",
      "date": "2026-07-05T00:00:00.000Z",
      "tags": [
        "ai-agents",
        "rl",
        "ml"
      ],
      "notes_markdown": "NVIDIA Technical Blog overview of training reasoning agents on multi-step environment trajectories.\n\n- Covers reward shaping, credit assignment over long tool horizons, and trajectory curation.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/06/hidden-technical-debt-agent-evaluation-infra",
      "slug": "2026/06/hidden-technical-debt-agent-evaluation-infra",
      "title": "Hidden Technical Debt of AI Systems: Agent Evaluation Infrastructure",
      "url": "https://leehanchung.github.io/blogs/2026/06/13/hidden-technical-debt-agent-evaluation-infra/",
      "externalUrl": "https://leehanchung.github.io/blogs/2026/06/13/hidden-technical-debt-agent-evaluation-infra/",
      "date": "2026-06-30T00:00:00.000Z",
      "tags": [
        "evals",
        "ai-agents",
        "software-engineering"
      ],
      "notes_markdown": "Lee Han Chung examining the operational friction of testing LLM agents in CI/CD.\n\n- Addresses sandbox state contamination, flaky non-deterministic tool mockings, and cost bloat.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/06/hamming-advice-on-research",
      "slug": "2026/06/hamming-advice-on-research",
      "title": "Dr. R. W. Hamming’s Advice on Research (\"You and Your Research\")",
      "url": "https://www.cs.utexas.edu/~dahlin/bookshelf/hamming.html",
      "externalUrl": "https://www.cs.utexas.edu/~dahlin/bookshelf/hamming.html",
      "date": "2026-06-29T00:00:00.000Z",
      "tags": [
        "research",
        "career"
      ],
      "notes_markdown": "Richard Hamming's legendary Bell Labs seminar on doing first-class research.\n\n- Why working on important problems, cultivating emotional courage, and communicating clearly define impactful scientific careers.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/06/are-we-ready-for-an-agent-native-memory-system",
      "slug": "2026/06/are-we-ready-for-an-agent-native-memory-system",
      "title": "Are We Ready For An Agent-Native Memory System?",
      "url": "https://arxiv.org/abs/2606.24775",
      "externalUrl": "https://arxiv.org/abs/2606.24775",
      "date": "2026-06-26T00:00:00.000Z",
      "tags": [
        "ai-agents",
        "systems",
        "research"
      ],
      "notes_markdown": "Proposes treating LLM agent memory analogous to OS virtual memory systems (registers, cache lines, paging, and persistent swap).\n\n- Evaluates hierarchical retrieval strategies for persistent long-horizon autonomy.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/06/lessons-from-research-scientist-job-search",
      "slug": "2026/06/lessons-from-research-scientist-job-search",
      "title": "Surprising Lessons from My Research Scientist Job Search",
      "url": "https://yongzx.github.io/blog/2026/06/24/job-search/",
      "externalUrl": "https://yongzx.github.io/blog/2026/06/24/job-search/",
      "date": "2026-06-25T00:00:00.000Z",
      "tags": [
        "career",
        "research"
      ],
      "notes_markdown": "Yong Zheng-Xin's candid breakdown of interviewing across industry AI research labs.\n\n- Covers job talk preparation, system design rounds, and choosing research teams.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/06/patterns-for-building-cybersecurity-evals",
      "slug": "2026/06/patterns-for-building-cybersecurity-evals",
      "title": "Patterns for Building Cybersecurity Evals",
      "url": "https://eugeneyan.com/writing/cybersecurity-evals/",
      "externalUrl": "https://eugeneyan.com/writing/cybersecurity-evals/",
      "date": "2026-06-25T00:00:00.000Z",
      "tags": [
        "evals",
        "ai-safety",
        "software-engineering"
      ],
      "notes_markdown": "Eugene Yan on designing automated evaluation environments for offensive and defensive AI security agents.\n\n- Emphasizes capturing exploit mechanics, deterministic scoring sandboxes, and defense-in-depth metrics.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/06/scaling-laws-carefully",
      "slug": "2026/06/scaling-laws-carefully",
      "title": "Scaling Laws, Carefully",
      "url": "https://lilianweng.github.io/posts/2026-06-24-scaling-laws/",
      "externalUrl": "https://lilianweng.github.io/posts/2026-06-24-scaling-laws/",
      "date": "2026-06-25T00:00:00.000Z",
      "tags": [
        "ml",
        "research",
        "llm"
      ],
      "notes_markdown": "Lilian Weng's thorough analysis comparing Chinchilla, Kaplan, and modern multi-modal scaling formulations.\n\n- Highlights subtle fitting caveats, vocabulary size adjustments, and inference-time compute scaling frontiers.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/06/lmsys-speculative-decoding-dflash-v2",
      "slug": "2026/06/lmsys-speculative-decoding-dflash-v2",
      "title": "The Next Generation of Speculative Decoding: DFlash and Spec V2",
      "url": "https://www.lmsys.org/blog/2026-06-15-next-generation-speculative-decoding-dflash-v2/",
      "externalUrl": "https://www.lmsys.org/blog/2026-06-15-next-generation-speculative-decoding-dflash-v2/",
      "date": "2026-06-15T00:00:00.000Z",
      "tags": [
        "llm",
        "systems",
        "developer-tools"
      ],
      "notes_markdown": "LMSYS technical update introducing diffusion-assisted token drafting and advanced tree verification algorithms delivering >3x decoding acceleration.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/06/opinionated-guide-to-ml-research",
      "slug": "2026/06/opinionated-guide-to-ml-research",
      "title": "An Opinionated Guide to ML Research",
      "url": "http://joschu.net/blog/opinionated-guide-ml-research.html",
      "externalUrl": "http://joschu.net/blog/opinionated-guide-ml-research.html",
      "date": "2026-06-15T00:00:00.000Z",
      "tags": [
        "research",
        "ml",
        "career"
      ],
      "notes_markdown": "John Schulman's personal reflections on problem selection, experimental iteration speed, and developing technical taste in machine learning.\n\n- Highly practical advice on research notebook discipline and avoiding premature optimization.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/06/principles-of-effective-research",
      "slug": "2026/06/principles-of-effective-research",
      "title": "Principles of Effective Research",
      "url": "https://michaelnielsen.org/blog/principles-of-effective-research/",
      "externalUrl": "https://michaelnielsen.org/blog/principles-of-effective-research/",
      "date": "2026-06-15T00:00:00.000Z",
      "tags": [
        "research",
        "career"
      ],
      "notes_markdown": "Michael Nielsen on developing scientific intuition, active note taking, spaced repetition for research mastery, and sustained deep exploration.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/06/survival-guide-to-a-phd",
      "slug": "2026/06/survival-guide-to-a-phd",
      "title": "A Survival Guide to a PhD",
      "url": "https://karpathy.github.io/2016/09/07/phd/",
      "externalUrl": "https://karpathy.github.io/2016/09/07/phd/",
      "date": "2026-06-15T00:00:00.000Z",
      "tags": [
        "research",
        "career"
      ],
      "notes_markdown": "Andrej Karpathy's timeless reflection on navigating graduate school, handling imposter syndrome, running fast experiments, and writing papers.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/05/luxical-lexical-dense-embeddings",
      "slug": "2026/05/luxical-lexical-dense-embeddings",
      "title": "Luxical: High-Speed Lexical-Dense Text Embeddings",
      "url": "https://www.datologyai.com/blog/introducing-luxical-embeddings",
      "externalUrl": "https://www.datologyai.com/blog/introducing-luxical-embeddings",
      "date": "2026-05-23T00:00:00.000Z",
      "tags": [
        "rag",
        "ml",
        "systems"
      ],
      "notes_markdown": "DatologyAI's architecture marrying inverted index lexical matching speeds with neural dense semantic embeddings in a unified vector index.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/05/theory-on-adam-instability-large-scale-ml",
      "slug": "2026/05/theory-on-adam-instability-large-scale-ml",
      "title": "A Theory on Adam Instability in Large-Scale Machine Learning",
      "url": "https://arxiv.org/abs/2304.09871",
      "externalUrl": "https://arxiv.org/abs/2304.09871",
      "date": "2026-05-21T00:00:00.000Z",
      "tags": [
        "ml",
        "research"
      ],
      "notes_markdown": "Explains why AdamW optimizer experiences sudden loss spikes when scaling model parameter counts.\n\n- Proposes variance bounds and learning rate schedule remedies.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/05/passion-for-software-in-age-of-ai",
      "slug": "2026/05/passion-for-software-in-age-of-ai",
      "title": "Passion for Software in the Age of AI",
      "url": "https://maxliani.wordpress.com/2026/05/11/passion-for-software-in-the-age-of-ai/",
      "externalUrl": "https://maxliani.wordpress.com/2026/05/11/passion-for-software-in-the-age-of-ai/",
      "date": "2026-05-20T00:00:00.000Z",
      "tags": [
        "software-engineering",
        "career"
      ],
      "notes_markdown": "Max Liani's meditation on why software craftsmanship, engineering taste, and deep systems understanding matter more than ever in the era of code generators.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/05/agent-evaluation-a-detailed-guide",
      "slug": "2026/05/agent-evaluation-a-detailed-guide",
      "title": "Agent Evaluation: A Detailed Guide",
      "url": "https://cameronrwolfe.substack.com/p/agent-evals",
      "externalUrl": "https://cameronrwolfe.substack.com/p/agent-evals",
      "date": "2026-05-19T00:00:00.000Z",
      "tags": [
        "evals",
        "ai-agents",
        "software-engineering"
      ],
      "notes_markdown": "Cameron Wolfe's extensive breakdown on constructing agent benchmarks.\n\n- Evaluates unit test assertions, trajectory fidelity, tool argument correctness, and human-in-the-loop rubrics.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/05/how-to-land-a-frontier-lab-job",
      "slug": "2026/05/how-to-land-a-frontier-lab-job",
      "title": "How to Land a Frontier Lab Job",
      "url": "https://vladfeinberg.com/2026/05/10/how-to-land-a-job-at-a-frontier-lab.html",
      "externalUrl": "https://vladfeinberg.com/2026/05/10/how-to-land-a-job-at-a-frontier-lab.html",
      "date": "2026-05-18T00:00:00.000Z",
      "tags": [
        "career",
        "systems"
      ],
      "notes_markdown": "Vlad Feinberg's concrete roadmap for preparing for systems and infrastructure roles at OpenAI, Anthropic, and DeepMind.\n\n- Focuses on CUDA profiling, distributed networking primitives, and hardware failure recovery.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/05/how-to-scale-your-model-jax-book",
      "slug": "2026/05/how-to-scale-your-model-jax-book",
      "title": "How To Scale Your Model (The JAX Scaling Book)",
      "url": "https://jax-ml.github.io/scaling-book/",
      "externalUrl": "https://jax-ml.github.io/scaling-book/",
      "date": "2026-05-18T00:00:00.000Z",
      "tags": [
        "systems",
        "ml",
        "developer-tools"
      ],
      "notes_markdown": "An exceptional open handbook on parallelizing transformers with JAX.\n\n- Deep dive into tensor parallelism (TP), pipeline parallelism (PP), Fully Sharded Data Parallel (FSDP), and overlap communication.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/05/apis-as-ladders",
      "slug": "2026/05/apis-as-ladders",
      "title": "APIs as Ladders",
      "url": "https://blog.sbensu.com/posts/apis-as-ladders/",
      "externalUrl": "https://blog.sbensu.com/posts/apis-as-ladders/",
      "date": "2026-05-15T00:00:00.000Z",
      "tags": [
        "developer-tools",
        "software-engineering"
      ],
      "notes_markdown": "Reflections on designing APIs that start simple for novices but gracefully unfurl full lower-level control as user needs deepen.\n\n- Great principles for framework and tool design.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/05/how-i-built-an-open-source-hinglish-tts",
      "slug": "2026/05/how-i-built-an-open-source-hinglish-tts",
      "title": "How I Built an Open-Source Hinglish TTS",
      "url": "https://harrrshall.github.io/hinglish-tts/",
      "externalUrl": "https://harrrshall.github.io/hinglish-tts/",
      "date": "2026-05-13T00:00:00.000Z",
      "tags": [
        "gen-ai",
        "developer-tools",
        "software-engineering"
      ],
      "notes_markdown": "Harshal's walkthrough of phonetic alignment, voice dataset curation, and code-switched Hindi-English speech synthesis.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/05/survey-on-policy-distillation-llms",
      "slug": "2026/05/survey-on-policy-distillation-llms",
      "title": "A Survey of On-Policy Distillation for Large Language Models",
      "url": "https://arxiv.org/abs/2604.00626",
      "externalUrl": "https://arxiv.org/abs/2604.00626",
      "date": "2026-05-13T00:00:00.000Z",
      "tags": [
        "distillation",
        "ml",
        "research"
      ],
      "notes_markdown": "Literature review outlining sequence-level distillation techniques across MiniLLM, GKD, and f-divergence generalizations.\n\n- Analyzes teacher guidance strategies during autoregressive sampling.",
      "site_url": "https://jayshah.dev/reads/"
    },
    {
      "id": "2026/05/sft-rl-on-policy-distillation-distributional-lens",
      "slug": "2026/05/sft-rl-on-policy-distillation-distributional-lens",
      "title": "SFT, RL, and On-Policy Distillation Through a Distributional Lens",
      "url": "https://nrehiew.github.io/blog/sft_rl_opd/",
      "externalUrl": "https://nrehiew.github.io/blog/sft_rl_opd/",
      "date": "2026-05-10T00:00:00.000Z",
      "tags": [
        "distillation",
        "rl",
        "ml"
      ],
      "notes_markdown": "A unified mathematical formulation connecting supervised fine-tuning (forward KL), RLHF (reverse KL with reward shaping), and on-policy distillation.\n\n- Clarifies why on-policy student generations prevent mode collapse and exposure bias.",
      "site_url": "https://jayshah.dev/reads/"
    }
  ]
}