Learning diary › Year 2025 › May, 2025 [2025-05]
Learning diary › Year 2025 › May, 2025 [2025-05]
2025-05-31 [2025-05-31]
2025-05-31 [2025-05-31]
#agent #dspy #lean #os #rust - found C++ to Rust Phrasebook - Simpler backoff - Consider Knitting - Lean-auto: An Interface between Lean 4 and Automated Theorem Provers - found Zipperposition - found Programming Beyond Practices - investigate DSPy - found Agenspy (Agentic DSPy) - found attachments - found Code Generation with AlphaCodium: From Prompt Engineering to Flow Engineering (code) - S*: Test Time Scaling for Code Generation - Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems
2025-05-30 [2025-05-30]
2025-05-30 [2025-05-30]
#build #formal #rust - found Grog: the mono-repo build tool for the grug-brained dev - then the configuration language Pkl: Configuration that is Programmable, Scalable, and Safe - The Grug Brained Developer (2022) - Rust doesn't actually follow its Golden Rule (when it comes to async functions) - Parse, don’t validate - Names are not type safety (2020) - When was the last time you broke production and how?
2025-05-29 [2025-05-29]
2025-05-29 [2025-05-29]
- Why Your AI Coding Assistant Keeps Doing It Wrong, and How To Fix It - workflows for ai coding - interesting take on human attention span - Compiling a Neural Net to C for a 1,744× speedup
2025-05-28 [2025-05-28]
2025-05-28 [2025-05-28]
#agent #codegen #gpu #os #proof #render #web - WebGPU Fluid Simulations: High Performance & Real-Time Rendering - learn about uzi - from LLM Codegen go Brrr – Parallelization with Git Worktrees and Tmux - A thought on JavaScript "proof of work" anti-scraper systems - What a Difference a Faster Hash Makes
2025-05-27 [2025-05-27]
2025-05-27 [2025-05-27]
- found Effekt Language: Home - The Three Kinds of Money - summarized with doubao, then inspired me to work on mermaid and markmap integration - The two types of open source
2025-05-26 [2025-05-26]
2025-05-26 [2025-05-26]
#bevy #exif #formal #id3 #news #os #physics #racket #rust #software - found F2, a bulk renaming tool that supports metadata such as EXIF, ID3 and more - Idiomatic Rust - Don't Worry About Lifetimes - Don't Unwrap Options: There Are Better Ways - Thinking in Expressions - found DumPy: NumPy except it's OK if you're dum - found einx - einx notation is inspired by einops, but introduces several novel concepts such as `[]`-bracket notation and full composability that allow using it as a universal language for tensor operations - Against Curry-Howard Mysticism - found Reinforcement Learning: An Overview - found Modeling Nonlinear Dynamics from Equations and Data — with Applications to Solids, Fluids, and Controls - found The Ultimate Guide to Fine-Tuning LLMs from Basics to Breakthroughs: An Exhaustive Review of Technologies, Research, Best Practices, Applied Research Challenges and Opportunities - #game - found Pong Wars - maybe it's fun to rewrite it in Zig, or with Bevy, possibly in 3D - Car Physics for Games - Games That Love To Take Things Away From You - hammurabi: A Rust recreation of the classic 1968 BASIC game - I made a font - Someone rewrote my Zig CLI in Rust? Let's compare! - found Transformer Copilot: Learning from The Mistake Log in LLM Fine-tuning - I've long wished for this kind of meaningful learning, instead of just fine-tuning weights according to loss - thoughts on software - Design Pressure: The Invisible Hand That Shapes Your Code - Systematic design of multi-join GROUP BY queries - Reinvent the Wheel - How to Guarantee Your Red Team Will Fail - On File Formats - What Works (and Doesn't) Selling Formal Methods
2025-05-25 [2025-05-25]
2025-05-25 [2025-05-25]
#formal #haskell #lean #ocaml #os #rust #zig - Having your compile-time cake and eating it too - Hindley-Milner (HM), the most popular human-friendly type system, used in Swift, Rust, Scala, Haskell, OCaml etc. - proposed four big features to replace pretty much all Rust macros, including derive ones, while keeping Rust's type system, and achieving (most of) Zig's comptime features - `@` or `@=` - `Abstract` - `TypeInfo` and `Field` - `Code` and `parse` - CAPTCHAs don't work any more for ticket sellers - Concepts vs type traits - tachy0n: The last 0day jailbreak - found Domain Theory Lecture Notes - generate and read the following DeepWiki: - ymndoseijin/zilliam - ShabbirHasan1/unsynn - AstatineAi/ocaml-forester - pygae/galgebra - utensil/forest - use GenSpark to generate Follow-up Research after Formalizing Geometric Algebra in Lean - Does gratitude increase happiness?
2025-05-24 [2025-05-24]
2025-05-24 [2025-05-24]
#agent #apl #cg #compiler #git #haskell #lean #os #rust #software - Anthropic’s ‘System Card’ for Claude 4 (Opus and Sonnet) - particularly fun reading - 4.1.1.1 Continuations of self-exfiltration attempts - 4.1.1.2 Opportunistic blackmail - 4.1.1.3 Self-exfiltration under extreme circumstances - 4.1.3 Excessive compliance with harmful system-prompt instructions - "mistakenly omitted the only finetuning dataset that included harmful system prompts" - Forgotten APL Influences (2016) - learn about Glean - a system for collecting, deriving and querying facts about source code - from Indexing Hackage: Glean vs. hiedb - Pain in the dots - different semantics of two and three dot notation for `git log` and `git diff` - Make Your Phone Grayscale - I've already started the practice, combining it with all-day nightshift mode - The GCC compiler backend can now fully bootstrap the Rust compiler
2025-05-23 [2025-05-23]
2025-05-23 [2025-05-23]
#ocaml #os #rust #tla - Why I love OCaml - How to Get Your Research Paper Accepted - Fork Union: Beyond OpenMP in C++ and Rust? - #lang - Wren, a small, fast, class-based concurrent scripting language - "Think Smalltalk in a Lua-sized package with a dash of Erlang and wrapped up in a familiar, modern syntax." - Talon: Write Raylib programs in Wren - Picat, a rule-based language, in which predicates, functions, and actors are defined with pattern-matching rules - from Finding hard 24 puzzles with planner programming - "the only language that I know has a built-in planner module" - Dark, a serverless backend language - Darklang Goes Open Source - Goodbye Dark Inc. - Hello Darklang Inc - Roto - from Introducing Roto: A Compiled Scripting Language for Rust - Janet #lisp - a more batteries-included Lua, with Lisp/Clojure-inspired syntax - see The smallest embeddable scripting language, part 1 - PEGs in a PEG - JNJ: J iN Janet - Bagatto, a static site generator written in Janet - goal, embeddable scripting array language - lamber, functional scripting language compiling to Lambda Calculus - found many papers, see some HF papers worth skimming
2025-05-22 [2025-05-22]
2025-05-22 [2025-05-22]
#context #formal #idea #os - Adventures in Symbolic Algebra with Model Context Protocol - maybe I should create MCP for GAlgebra - followup: Interfacing MCP with Combinatorial, Convex, and SMT Solvers - Transpiler is a meaningless word (2023) - maybe that means I should give up on the idea that I could build a Forester transpiler - found Obsidian Bases
2025-05-21 [2025-05-21]
2025-05-21 [2025-05-21]
#os #sqlite - If composers were hackers - found Litestream: Revamped, making SQLite apps reliably recoverable from object storage - The Ingredients of a Productive Monorepo
2025-05-20 [2025-05-20]
2025-05-20 [2025-05-20]
#gpu #news #os #web - Not causal chains, but interactions and adaptations - RCA(root cause analysis) model v.s. RE(resilience engineering) model, for incidents - related - The same incident never happens twice, but the patterns recur over and over - Quick takes on the GCP public incident write-up - “What went well” is more than just a pat on the back - Behind the scenes: Redpanda Cloud’s response to the GCP outage (on HN) (on lobste.rs) - Implicit is better than explicit - Open Source Can't Coordinate - Particle Life simulation in browser using WebGPU
2025-05-19 [2025-05-19]
2025-05-19 [2025-05-19]
#agent #docker #lemmy #mastodon #os - Agent Recursion - Semgrep: AutoFixes using LLMs - found a few alternatives to `semgrep`, e.g. ast-grep - Adding Mastodon Comments to your Blog - Mastodon as comment system for your static blog - My blog now has Lemmy comments - learn about Glasp: Highlight the Internet & Build Your AI Clone - async/await versus the Calloop Model - Mimalloc Cigarette: Losing one week of my life catching a memory leak - Ditching Obsidian and building my own - found Qwen3 Technical Report - found TestDesiderata - found macOS VMs in a Docker Container
2025-05-18 [2025-05-18]
2025-05-18 [2025-05-18]
#apl #compiler #game #gpu #os #physics #quantum #shader #zig - found Algorithms by Jeff Erickson (2019) - Comparing parallel functional array languages: Programming and performance[van2025comparing] - learned that APL can run on GPU via Dyalog, the paper even discussed a flash attention implementation - BQN doesn't seem to run on GPU - it might be interesting to use arrayfire C API, even on Mac - prefer tilelang to DaCe - found A JavaScript library for building parsers, interpreters, compilers, and more - Layers All The Way Down: The Untold Story of Shader Compilation - introduces the motivation behind MojoShader, from Direct3D HLSL bytecode or assembly source to other shader languages - found Mystical, a programming language that resembled magical circles - Bare Metal Zig (2023) - Google Scholar is manipulatable - found Path Integrals in Quantum Mechanics, Statistics, Polymer Physics, and Financial Markets by Hagen Kleinert
2025-05-17 [2025-05-17]
2025-05-17 [2025-05-17]
#elixir #game #ocaml #os #rust #software #web #zig - Parallel Scaling Law for Language Models - #lang - The Language That Never Was - Leaving Rust gamedev after 3 years - Mun - jank - The Little Book of Rust Macros - MonoGame - #rust/error - Using unwrap() in Rust is Okay - anyhow for applications and thiserror for libraries - On Error Handling in Rust - error_set, inspired by Zig's error set - snafu - found GNU Parallel Examples - found Lua for Elixir - found OCaml Web Development: Essential Tools and Libraries in 2025 - found Teal: a statically-typed dialect of Lua, a statically-typed dialect of Lua
2025-05-16 [2025-05-16]
2025-05-16 [2025-05-16]
#formal #game #news #os #rust #tla #zig - I don’t like NumPy - Writing a better code with pytorch and einops - Einops and Einsum Summarized - looking for Zig way to implement eiops - Interesting (mis-)use cases for comptime - zig-infix-parser - Zig Comptime is amazing - ZEIN, Zig-based implementation of general-rank tensors - Functional Programming in Zig - What is Zig's Comptime? - comath: comptime math, used by zilliam - alg, Algebra for Zig - mecha, A parser combinator library for Zig - from Zig, Parser Combinators - and Why They're Awesome - parcom, see also this blog post - einops in Rust, with only backend `tch` - Einsums and einops-cpp are in C++ - The current state of TLA⁺ development - found Spectacle and tlafmt - 10 Years of Stable Rust: An Infrastructure Story - formal methods projects built on Rust - Beyond 'Aha!': Toward Systematic Meta-Abilities Alignment in Large Reasoning Models - Declaring a friendship to self - Internet Artifacts - discovered Helicopter game
2025-05-15 [2025-05-15]
2025-05-15 [2025-05-15]
#formal #forth #os #tla #zig - Introducing oniux: Kernel-level Tor isolation for any Linux app - code - found smoltcp - The cryptography behind passkeys - How fast is CeTZ-Plot? - ZJIT has been merged into Ruby - Instead of compiling YARV bytecode directly to the low-level IR (LIR), it uses an high-level SSA-based intermediate representation (HIR) - In the bytecode, which is tersely encoded, jumps are offsets, some control-flow is implicit, and most dataflow is via the stack. - By contrast, HIR looks more like a graph. Jumps have pointers to their targets and there’s no stack: instructions that use data have pointers directly to the instructions that create the data. - found Warteschlangensimulator, a event-driven, stochastic simulator for queueing systems - found kowk, a ~100k-pod k8s cluster simulator - Data is code, about philosophy behind Forth - also What the hell is Forth? - found Insights into DeepSeek-V3: Scaling Challenges and Reflections on Hardware for AI Architectures - found Marigold: Affordable Adaptation of Diffusion-Based Image Generators for Image Analysis - Marigold Computer Vision - Zig App Release and Updates via Github ⚡
2025-05-14 [2025-05-14]
2025-05-14 [2025-05-14]
#agent #ebpf #os #physics #rust #software #tui #web #yaml - try genspark - impressive result - one deep research per day for free - Read the Code, Not the Profile - Dramatic percentages in software is just linear improvement - Writing that changed how I think about PL - Experiment on your code freely with Git worktree - Lock-Free Rust: How to Build a Rollercoaster While It’s on Fire - found - Spall: a code profiler that runs in your browser - TUI for search & replace - scooter: Interactive find and replace in the terminal - serpl: A simple terminal UI for search and replace, ala VS Code. - scooter wins as it supports filtering files, both support regex and replace preview - Avian Physics 0.3 - Makepad 1.0 : Rust UI Framework - checked out YS — YAML Done Wisely - I wish to use it as a template engine, but the scripting syntax is not delicious - Two months in Servo: CSS nesting, Shadow DOM, Clipboard API, and more - Servo is still very actively growing - Misadventures in DTrace: how to debug the macOS kernel - I need to compare more with eBPF and DTrace - found DTrace book
2025-05-12 [2025-05-12]
2025-05-12 [2025-05-12]
#apl #formal #os #proof #rust #sec #tla #web - Are We Serious About Using TLA+ For Statistical Properties? - found FizzBee for behavior correctness verification and performance modelling, based on simulation - I've wished for a tool like this for a long time - see also Modular verification of MongoDB Transactions using TLA+ - Flattening Rust's Learning Curve - The best open source project for someone might not be yours, and that's OK - A review of documentation in the Rust ecosystem - found Blessed.rs: Recommended Crate Directory - Secure by Design: Google’s Perspective on Memory Safety (2024) - Microservices Are a Tax Your Startup Probably Can’t Afford - Private Internet (2024) - A tool to verify estimates, II: a flexible proof assistant
2025-05-09 [2025-05-09]
2025-05-09 [2025-05-09]
#os #rust #zig - Memory Safety Features in Zig - very well summarized, with great examples - Reservoir Sampling - You can use C-Reduce for any language - Rust Dependencies scare Me - a comment reminds me of Build It Yourself - What's in an e-graph? - A catalog of ways to generate SSA
2025-05-08 [2025-05-08]
2025-05-08 [2025-05-08]
#agent #gpu #idea #interop #os #software #web #zig - Why transformers need adam: A hessian perspective[zhang2024transformers] - insightful, and well summarized related work - trying to figure out a way to let AI agent to read all papers citing an paper, and write a summary of the follow-up research - The magic of software; or, what makes a good engineer also makes a good engineering organization - "Engineers are most capable and most effective when abstraction layers serve as shorthand for an understanding of what it is doing on your behalf, rather than a black box" - "start with deep understanding as the basis for innovation – cultivating the curiosity to look inside the black boxes" - "The magic of both software and software organizations comes from those moments where insight into how something works sparks entirely new ideas about what it could become" - related: Reinvent the Wheel - Implement your language twice - Futhark is a statically typed, data-parallel, and purely functional array language with AoT compilation to CUDA, OpenCL, HIP, and WebGPU (WIP) - also Comparing the performance of OpenCL, CUDA, and HIP - Objective-C interop with Zig? - got zig to work inside lima with minimal setup
2025-05-07 [2025-05-07]
2025-05-07 [2025-05-07]
#os #rust #tui #web #zig - Zig’s Low-Level Safety Features Leave Rust in the Dust - Zig defer Patterns - Glossary Web Component - Recreating an iOS Animation with GLSL (interactive tutorial) - found Argo CD - Declarative Continuous Delivery for k8s - found bash/POSIX-compatible shell implemented in Rust - found nnd: A TUI alternative to gdb
2025-05-07 [2025-05-07-2]
2025-05-07 [2025-05-07-2]
#ebpf #os #rust #tui #zig - Zig’s Low-Level Safety Features Leave Rust in the Dust - found Parallel, Concurrent and Distributed Programming in Scala - found nnd: A TUI alternative to gdb - found A part of the fundamental group of an n-dimensional beanA - Building a regex engine in C++ - wondering about chaos engineering in Zig with eBPF #ebpf - learn more about zbpf - maybe it's easiest to use lima to test eBPF - found Pixie when reading Debugging Production: eBPF Chaos - Introducing Minderbinder which is in Go - Implementing fast TCP fingerprinting with eBPF - using Cilium in Go - ebpf series - Unnamed Memories - Introducing TCP-in-UDP solution (eBPF) - Go Allocation Probe (on HN) - KernelScript eBPF-centric programming language (on HN) #lang - OOMProf - Take a heap profile just before OOMkill using eBPF #chaos - found alternative open source front-ends for popular internet platforms
2025-05-06 [2025-05-06]
2025-05-06 [2025-05-06]
#agent #benchmark #blogging #context #formal #game #gpu #news #optimization #os - LM - survey papers - A Survey on Inference Engines for Large Language Models: Perspectives on Optimization and Efficiency - Low-Precision Training of Large Language Models: Methods, Challenges, and Opportunities - sadly, NormalFloat used by QLoRA is excluded from the paper, as it is "specifically designed for pretrained fixed parameters, which only participate in the forward inference stage during training." - improvements on reasoning - Think on your Feet: Adaptive Thinking via Reinforcement Learning for Social Agents - the model is trained to switch between different complexity level of thinking mode based on context for an optimal balance between correct answer and reasoning length - the thinking modes are inspired by Hierarchical Cognitive Control Theory (HCCT) - RM-R1: Reward Modeling as Reasoning - provides good insight into reward modeling and how each training stage boosts the model's performance - other - Unlearning Sensitive Information in Multimodal LLMs: Benchmark and Attack-Defense Evaluation - I'm still interested in unlearning and model editing - Benchmarking Crimes Meet Formal Verification - learn about Systems Benchmarking Crimes - #zig - Implementing a Struct of Arrays - Zig has easy-to-use support of SoA via MultiArrayList - learn about Data-Oriented Design - I can't help but feel that new features of C++ such as reflection is really becoming syntax cancer - Reflecting on a year of Gamedev in Zig - Jai, the game programming contender - bringing data to AI, not the other way around - The Data Surrender Trap: How Enterprises Are Losing Control in the AI Gold Rush—and the Simple Fix - lifestyle - Protect The Habit - skimmed blogging in isolation - skimmed modern-latex: A short guide to LaTeX that avoids legacy cruft - The Beauty Of Having A Pi-hole - own your own DNS - An appeal to Apple from Anukari regarding GPU frequency scaling - I'm amazed that the author is using GPU to simulate physics for realtime audio
2025-05-04 [2025-05-04]
2025-05-04 [2025-05-04]
#git #interop #lean #os - found pyonji, a tool to support sr.ht style e-mail patches - Git: programmatic staging - and learn about `grepdiff`, unfortunately, it's not available on Mac - `git add -p` is also acceptable for a small number of hunks - MathML with Pandoc - Starting on seamless C++ interop in jank - found Anemll: Artificial Neural Engine Machine Learning Library
2025-05-02 [2025-05-02]
2025-05-02 [2025-05-02]
#formal #os #web - DeepSeek-prover-V2: Advancing formal mathematical reasoning via reinforcement learning for subgoal decomposition[ren2025deepseekproverv2] - A survey on post-training of large language models[tie2025survey] - notes on LM could be based on this survey and the following papers related to r1 - 100 days after DeepSeek-R1: A survey on replication studies and more directions for reasoning language models[zhang2025days] - Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning[guo2025deepseek] - should revisit - found critics of r1/GRPO - Understanding r1-zero-like training: A critical perspective[liu2025understanding] - Does reinforcement learning really incentivize reasoning capacity in LLMs beyond the base model?[yue2025does] - Kimina-prover preview: Towards large formal reasoning models with reinforcement learning[wang2025kimina] - found A Survey of Interactive Generative Video - Polishing your typography with line height units - Solving Sudoku with Algebraic Geometry and Computer Algebra : A C Programming Approach