Random thoughts and projects.https://tylercrosse.com/https://tylercrosse.com/ideas/2020/git-stash/https://tylercrosse.com/ideas/2020/git-stash/Quick tips for using git stash.Sun, 30 Aug 2020 00:00:00 GMThttps://tylercrosse.com/ideas/2021/nand2tetris-part1/https://tylercrosse.com/ideas/2021/nand2tetris-part1/The first half of the Nand2Tetris course: Boolean logic, arithmetic, memory, machine language, CPU architecture, and an assembler. We build the Hack computer up from Nand gates.Sun, 26 Sep 2021 00:00:00 GMThttps://tylercrosse.com/ideas/2022/exponential-backoff/https://tylercrosse.com/ideas/2022/exponential-backoff/How to automaically retry requests and not blow stuff upWed, 28 Sep 2022 00:00:00 GMThttps://tylercrosse.com/ideas/2022/nand2tetris/https://tylercrosse.com/ideas/2022/nand2tetris/Building a modern computer from first principles — a two-part retrospective on the Nand2Tetris courseSat, 12 Mar 2022 00:00:00 GMThttps://tylercrosse.com/ideas/2022/nand2tetris-vm-translator/https://tylercrosse.com/ideas/2022/nand2tetris-vm-translator/The start of the second half of Nand2Tetris: stack machines, virtual memory segments, and a VM translator that compiles VM bytecode down to Hack assembly.Sat, 12 Mar 2022 00:00:00 GMThttps://tylercrosse.com/ideas/2022/planting-seeds/https://tylercrosse.com/ideas/2022/planting-seeds/Planting the seeds of my own digital gardenFri, 11 Mar 2022 00:00:00 GMThttps://tylercrosse.com/ideas/2022/python-bitwise-set-and-dicts/https://tylercrosse.com/ideas/2022/python-bitwise-set-and-dicts/Bitwise or getting a bit wise? Quick post on quirky Python syntax for working with sets and dicts.Thu, 27 Oct 2022 00:00:00 GMThttps://tylercrosse.com/ideas/2022/search-engine-fundamentals/https://tylercrosse.com/ideas/2022/search-engine-fundamentals/Essential terms and concepts of how search works. Queries, indexes, & relevance.Sun, 20 Nov 2022 00:00:00 GMThttps://tylercrosse.com/ideas/2023/language-models/https://tylercrosse.com/ideas/2023/language-models/From n-grams to ChatGPT, how language models work and how they can be used to solve real-world problems.Sun, 21 May 2023 00:00:00 GMThttps://tylercrosse.com/ideas/2023/semantic-search/https://tylercrosse.com/ideas/2023/semantic-search/What's an embedding vector, and how can we use neural networks to improve the relevance of search results?Wed, 16 Aug 2023 00:00:00 GMThttps://tylercrosse.com/ideas/2023/tech-debt/https://tylercrosse.com/ideas/2023/tech-debt/Tech debt, knowledge debt, and the kitchen sink.Tue, 27 Jun 2023 00:00:00 GMThttps://tylercrosse.com/ideas/2025/ai-safety-evals-paper-club-notes/https://tylercrosse.com/ideas/2025/ai-safety-evals-paper-club-notes/My notes and takeaways from the BlueDot AI Safety Evals paper club, covering recent papers on AI alignment, security, and evaluations.Tue, 09 Sep 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/debunking-the-cuda-myth/https://tylercrosse.com/ideas/2025/debunking-the-cuda-myth/A written version of my CS 8803 paper presentation on Lee et al.'s ISCA 2025 evaluation of Intel's Gaudi-2 NPU: performance, programmability, and whether CUDA is really the moat.Sat, 01 Nov 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/gios-pr1/https://tylercrosse.com/ideas/2025/gios-pr1/A project that explores multithreading, sockets, and the basics of operating systemsMon, 23 Jun 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/gios-pr3/https://tylercrosse.com/ideas/2025/gios-pr3/Exploring Inter-Process Communication and synchronization in C by building a high-performance proxy and cache.Thu, 10 Jul 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/gios-pr4/https://tylercrosse.com/ideas/2025/gios-pr4/An exploration of remote communication, data consistency, and synchronization in a C++ distributed file system project.Tue, 22 Jul 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/gios-retro/https://tylercrosse.com/ideas/2025/gios-retro/What I Learned in Graduate Introduction to Operating Systems (GIOS) - A RetrospectiveSun, 22 Jun 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/hierarchical-reasoning-models/https://tylercrosse.com/ideas/2025/hierarchical-reasoning-models/A written version of my Latent Space talk on HRM: how the architecture works, and what the ARC Prize analysis says actually drives its performance.Wed, 03 Sep 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/icml/https://tylercrosse.com/ideas/2025/icml/Some notes and observations from my time at the 2025 International Conference for Machine Learning (ICML)Wed, 30 Jul 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/language-diffusion-survey/https://tylercrosse.com/ideas/2025/language-diffusion-survey/A written version of my Latent Space Paper Club talk surveying diffusion models for language, from denoising autoencoders and DDPM through LLaDA, Block Diffusion, and d1.Wed, 28 May 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/ml-a1/https://tylercrosse.com/ideas/2025/ml-a1/A conceptual walkthrough of the supervised learning process, covering data analysis, model evaluation, and the fundamental trade-offs involved in building effective classifiers.Sun, 10 Aug 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/ml-research-notes/https://tylercrosse.com/ideas/2025/ml-research-notes/A summary of my takeaways from three influential articles on conducting effective empirical AI alignment research.Sat, 18 Oct 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/ml-retro/https://tylercrosse.com/ideas/2025/ml-retro/A survey of concepts covered in my graduate Machine Learning courseSat, 09 Aug 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/mscs/https://tylercrosse.com/ideas/2025/mscs/What I studied and built during Georgia Tech's MSCS program in machine learning and computing systemsFri, 20 Jun 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/parallelism-concurrency/https://tylercrosse.com/ideas/2025/parallelism-concurrency/A dive into concurrency and parallelism across different levels of computer systemsTue, 29 Jul 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2025/uv/https://tylercrosse.com/ideas/2025/uv/An introduction to uv, a fast, Rust-based tool for Python packagingFri, 17 Oct 2025 00:00:00 GMThttps://tylercrosse.com/ideas/2026/calorie-tracker/https://tylercrosse.com/ideas/2026/calorie-tracker/A mobile-first nutrition PWA that estimates calories and macros from food photos, scans barcodes and menus, and turns model output into a reviewable first draft.Sat, 27 Jun 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/circuits-attention-qk/https://tylercrosse.com/ideas/2026/circuits-attention-qk/Notes on QK attributions, attention-head loadings, and what feature interactions reveal about why a transformer attends to one token rather than another.Sun, 19 Jul 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/circuits-attribution-graphs/https://tylercrosse.com/ideas/2026/circuits-attribution-graphs/Notes on cross-layer transcoders, local replacement models, and attribution graphs — and the case-study companion paper on Claude 3.5 Haiku.Sat, 18 Jul 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/circuits-crosscoders/https://tylercrosse.com/ideas/2026/circuits-crosscoders/Notes on sparse crosscoders: one shared feature dictionary across layers or across models, and what that enables for cross-layer analysis and model diffing.Sat, 18 Jul 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/circuits-framework/https://tylercrosse.com/ideas/2026/circuits-framework/Notes on the 2021 paper that reframes transformers as operations on a shared residual stream: QK/OV circuits, virtual weights, and path decompositions of attention-only models.Sun, 12 Jul 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/circuits-induction-heads/https://tylercrosse.com/ideas/2026/circuits-induction-heads/Notes on induction heads, the training phase change in which they form, and the six-part argument connecting them to in-context learning.Sun, 12 Jul 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/circuits-monosemanticity/https://tylercrosse.com/ideas/2026/circuits-monosemanticity/Notes on decomposing MLP activations into interpretable features with sparse autoencoders, from a one-layer transformer to Claude 3 Sonnet.Tue, 14 Jul 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/circuits-superposition/https://tylercrosse.com/ideas/2026/circuits-superposition/Notes on how sparse features fit into superposition, the geometry and interference that result, and a toy-model account of memorization and double descent.Tue, 14 Jul 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/git-worktrees/https://tylercrosse.com/ideas/2026/git-worktrees/Work on multiple branches at once without ever touching git stash.Thu, 30 Apr 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/gpu-p1p2/https://tylercrosse.com/ideas/2026/gpu-p1p2/Writing real GPU kernels, exploring shared memory tiling, parallel sorting algorithms, and performance optimization on an NVIDIA H100.Wed, 25 Feb 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/gpu-p3p4/https://tylercrosse.com/ideas/2026/gpu-p3p4/Building a cycle-level GPU simulator from the inside, implementing GTO and CCWS warp schedulers, modeling compute instruction latencies, and extending to tensor core simulation.Wed, 25 Feb 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/gpu-p5/https://tylercrosse.com/ideas/2026/gpu-p5/A simple if statement can halve GPU throughput, exploring how static dataflow analysis detects branch divergence by building a def-use chain over a SASS control flow graph.Wed, 25 Feb 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/gpu-p6/https://tylercrosse.com/ideas/2026/gpu-p6/Writing a FlashAttention CUDA kernel from scratch, tiling the attention matrix to avoid materializing N×N memory, building a KV cache for token generation, and running GPT-2 with custom kernels end-to-end.Wed, 25 Feb 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/gpu-retro/https://tylercrosse.com/ideas/2026/gpu-retro/What I learned in GPU Hardware and Software (CS 8803).Wed, 25 Feb 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/llm-morality-mech-interp/https://tylercrosse.com/ideas/2026/llm-morality-mech-interp/How moral fine-tuning on iterated prisons changes LLMs.Sat, 28 Feb 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/paper-asymmetric-goal-drift/https://tylercrosse.com/ideas/2026/paper-asymmetric-goal-drift/Coding agents violate system-prompt constraints more readily when the constraint opposes strongly held trained-in values like security and privacy.Thu, 16 Apr 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/paper-attack-selection/https://tylercrosse.com/ideas/2026/paper-attack-selection/Control evaluations that let the red team choose when to attack produce meaningfully lower safety estimates than evaluations that ignore attack selection.Thu, 25 Jun 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/paper-inherited-goal-drift/https://tylercrosse.com/ideas/2026/paper-inherited-goal-drift/Frontier agents resist direct pressure to abandon their goals, but drift when conditioned on flawed trajectories inherited from weaker agents.Thu, 16 Apr 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/paper-offline-selectors/https://tylercrosse.com/ideas/2026/paper-offline-selectors/A diagnostic study of why learned model selectors fail to recover a large per-instance oracle gap on edX dropout prediction.Thu, 25 Jun 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/paper-same-facts-different-updates/https://tylercrosse.com/ideas/2026/paper-same-facts-different-updates/Whether an LLM sees its own prior response changes how it updates on identical new information in medical allocation scenarios.Thu, 25 Jun 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/persona-vectors/https://tylercrosse.com/ideas/2026/persona-vectors/Notes on two accounts of character in language models: persona vectors, which locate traits as steerable directions in activation space, and the persona selection model, which explains why such directions exist.Mon, 20 Jul 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/real-time-collaboration-contention/https://tylercrosse.com/ideas/2026/real-time-collaboration-contention/A survey of last-writer-wins, operational transformation, CRDTs, optimistic concurrency, and locks, and what each one does when two people edit the same thing at once.Wed, 08 Jul 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/transformer-circuits/https://tylercrosse.com/ideas/2026/transformer-circuits/Paper-by-paper notes on Anthropic's Transformer Circuits thread: core claims, methods in plain language, and what the evidence does and doesn't show.Sun, 12 Jul 2026 00:00:00 GMThttps://tylercrosse.com/ideas/2026/usage-bar/https://tylercrosse.com/ideas/2026/usage-bar/Extending an MIT-licensed menu bar app to show Claude Code and OpenAI Codex usage side by sideSat, 27 Jun 2026 00:00:00 GMT