Blog Categories
agents (2)
ai (3)
ai-agents (19)
ai-tools (2)
android (1)
audio (1)
benchmarks (2)
cli-tools (10)
compilers (7)
computer-history (1)
data-formats (1)
deep-dive (9)
developer-workflow (1)
dogfooding (1)
efficiency (2)
emacs (3)
embedded (8)
embedded-systems (3)
emulators (1)
explainers (35)
forth (5)
fpga (1)
fundamentals (1)
games (3)
gpu (2)
graphics (3)
hardware (5)
homelab (1)
interpretability (1)
introduction (1)
language-design (18)
languages (37)
llm (49)
machine-learning (76)
math (1)
mcp (1)
meta (1)
mobile (1)
multi-agent (1)
operating-systems (1)
personal (6)
process (3)
productivity (1)
programming (1)
programming-history (10)
programming-languages (10)
projects (12)
rabbit-hole (8)
research (22)
retrocomputing (7)
rust (35)
security (1)
self-supervised-learning (1)
smalltalk (1)
systems (3)
tbt (13)
testing (2)
tools (30)
vibe-coding (10)
visual-programming (1)
visualization (1)
web (1)
webassembly (2)
agents
- AI Tools #4: Pi --- The Minimal Agent That Stays Out of the Way
- AI Tools #5: nono --- Sandboxing Pi Without Breaking the Loop
ai
- Saw #1: pjmai-rs, Rig, and langchain-rust
- Lucy 20%: Upgrading My Home AI Cluster
- MCP: Teaching Claude to Play (and Trash Talk)
ai-agents
- AI Tools #7: A Coding Agent Small Enough to Understand
- AI Tools #6: Evaluating Local Models in a Plan-Execute-Review Loop
- AI Tools #2: AgentRail Mid-Saga --- Insert, Reorder, Reopen, Recover
- Saw #7: Prolog, Many-Agent Isolation, Self-Hosting Assembler, and MLPL
- Saw #6: Agent Coordination, Fuzzing Tests, Vendoring, and Emacs Graphics
- Saw #5: Sagas, Languages, and Compiler Chains
- Saw #4: All Together Now --- Emacs Meets the Multi-Agent Orchestra
- Saw #3: agentrail-rs --- From Walking Skeleton to Dual Memory
- AI Tools #1: XSkill --- A Memory Layer for Multimodal Agents
- How AI Learns Part 7: Designing a Continuous Learning Agent
- How AI Learns Part 6: Toward Continuous Learning
- How AI Learns Part 5: Context Engineering & Recursive Reasoning
- How AI Learns Part 4: Memory-Based Learning
- How AI Learns Part 3: Weight-Based Learning
- How AI Learns Part 2: Catastrophic Forgetting vs Context Rot
- How AI Learns Part 1: The Many Meanings of Learning
- music-pipe-rs: Unix Pipelines for MIDI Composition
- midi-cli-rs: Extending with Custom Mood Packs
- midi-cli-rs: Music Generation for AI Coding Agents
ai-tools
- Personal Software #9: Vibe-Maintenance --- When AI Agents Don't Just Write Code, They Fix Bugs
- AI Tools #3: sw-checklist --- Reining In AI Coding Agents With a Code-Metrics Ratchet
android
audio
benchmarks
cli-tools
- Personal Software #9: Vibe-Maintenance --- When AI Agents Don't Just Write Code, They Fix Bugs
- Personal Software #8: One Ring to Rule Them All --- sw-launcher's Memory Profiles, Heap Budgets, and a Working Scenario A
- AI Tools #3: sw-checklist --- Reining In AI Coding Agents With a Code-Metrics Ratchet
- AI Tools #2: AgentRail Mid-Saga --- Insert, Reorder, Reopen, Recover
- Saw #4: All Together Now --- Emacs Meets the Multi-Agent Orchestra
- Saw #3: agentrail-rs --- From Walking Skeleton to Dual Memory
- TBT #7: reg-rs - Regression Testing from C++ to Java to Rust
- pjmai-rs: Navigation History and Fuzzy Completion
- rank-wav: Ranking Audio Files by Acoustic Quality
- TBT #6: PJMAI-RS - A Shell That Knows Your Projects
compilers
- TBT #10: Mass Compile and PL/EDIT --- 1980s Productivity Tools, Reborn on COR24 PL/SW
- Bucket List #3: 3D Source Code, Five New Languages, and Visible Compilers
- Dogfooding #1: YAGNI Until You Do --- A Pascal P-Code Bump Allocator Grows a Free List
- Saw #8: Tuplet, Smalltalk-on-BASIC, proto-forth-wasm, sw-MLPL Split, and I2C on COR24
- Saw #7: Prolog, Many-Agent Isolation, Self-Hosting Assembler, and MLPL
- Saw #6: Agent Coordination, Fuzzing Tests, Vendoring, and Emacs Graphics
- Saw #5: Sagas, Languages, and Compiler Chains
computer-history
data-formats
deep-dive
- Rabbit-hole #9: X_eTaL and Mojo on a Qualcomm Adreno 702
- Rabbit-hole #8: From X_eTaL to the Metal, One Kernel Down Every GPU Toolchain
- Rabbit-hole #7: The Phoenix Combinator, and Why X_eTaL Brackets Its Forks
- Rabbit-hole #6: The Sage Bird --- Y Combinators in an Eager Array Language
- Rabbit-hole #5: FORTH --- Retargetable Codegen and the Forward-Reference Problem
- Rabbit-hole #4: FORTH --- Dictionary Compaction and Specialized Images
- Rabbit-hole #3: FORTH --- Life After Hashing
- Rabbit-hole #2: FORTH --- FIND and the Cost of a Name
- Embedded #3: How Much of Forth Can Be Forth? A Kernel Self-Hosting Spectrum
developer-workflow
dogfooding
efficiency
emacs
- Saw #8: Tuplet, Smalltalk-on-BASIC, proto-forth-wasm, sw-MLPL Split, and I2C on COR24
- Saw #6: Agent Coordination, Fuzzing Tests, Vendoring, and Emacs Graphics
- Saw #4: All Together Now --- Emacs Meets the Multi-Agent Orchestra
embedded
- 256 Bytes to 640 Gigabytes: Fifty Years of Memory, in Constant Dollars
- Saw #13: A Second Host in the Cloud, a Word Cloud, and Three Benches to Set Up
- Made Visible #1: A Microkernel's Storage, Drawn
- Saw #9: Espanso, Kate, ShareX, and Pluggable I2C Devices on COR24
- Personal Software #9: Vibe-Maintenance --- When AI Agents Don't Just Write Code, They Fix Bugs
- Personal Software #8: One Ring to Rule Them All --- sw-launcher's Memory Profiles, Heap Budgets, and a Working Scenario A
- Dogfooding #1: YAGNI Until You Do --- A Pascal P-Code Bump Allocator Grows a Free List
- Embedded #3: How Much of Forth Can Be Forth? A Kernel Self-Hosting Spectrum
embedded-systems
- Embedded #2: COR24-RS --- Learn Assembly in Your Browser
- Embedded #1: BMP280 Driver --- From Prototype to Patent Proof-of-Concept
- Rabbit-hole #1: Poor Man's Rust-to-Unsupported-ISA Translator
emulators
explainers
- ML Frontier #05: Grokking --- Delayed Generalization
- ML Frontier #04: Is Chain of Thought Real?
- ML Frontier #03: Structure Beats Scale --- Knowledge Graphs and Domain-Specific Superintelligence
- ML Frontier #02: In-Context Reinforcement Learning
- ML Frontier #01: Neural Collapse
- Five ML Concepts - #30: The Journey So Far
- Five ML Concepts - #29
- Five ML Concepts - #28
- Five ML Concepts - #27
- Five ML Concepts - #26
- Five ML Concepts - #25
- Five ML Concepts - #24
- Five ML Concepts - #23
- Five ML Concepts - #22
- Five ML Concepts - #21
- Five ML Concepts - #20
- Five ML Concepts - #19
- Five ML Concepts - #18
- Five ML Concepts - #17
- Five ML Concepts - #16
- Five ML Concepts - #15
- Five ML Concepts - #14
- Five ML Concepts - #13
- Five ML Concepts - #12
- Five ML Concepts - #11
- Five ML Concepts - #10
- Five ML Concepts - #9
- Five ML Concepts - #8
- Five ML Concepts - #7
- Five ML Concepts - #6
- Five ML Concepts - #5
- Five ML Concepts - #4
- Five ML Concepts - #3
- Five ML Concepts - #2
- Five ML Concepts - #1
forth
- Saw #8: Tuplet, Smalltalk-on-BASIC, proto-forth-wasm, sw-MLPL Split, and I2C on COR24
- Rabbit-hole #5: FORTH --- Retargetable Codegen and the Forward-Reference Problem
- Rabbit-hole #4: FORTH --- Dictionary Compaction and Specialized Images
- Rabbit-hole #3: FORTH --- Life After Hashing
- Rabbit-hole #2: FORTH --- FIND and the Cost of a Name
fpga
fundamentals
games
- Array Languages #6: X_eTaL-games, Rules That Are Arrays
- Voxels in an Array Language: A 3D World in Dyalog APL and X_eTaL
- Personal Software #10: A Sudoku That Teaches You the Strategies
gpu
- Array Languages #10: X_eTaL on Metal and CUDA, Through Mojo
- Array Languages #7: X_eTaL-gpu, Array Programs on GPUs Old and New
graphics
- Array Languages #9: X_eTaL and Blender, Scenes and Animation from Arrays
- Array Languages #4: X_eTaL-extensions, Extending the Machine
- Voxels in an Array Language: A 3D World in Dyalog APL and X_eTaL
hardware
- Array Languages #8: X_eTaL-fpga, Array Programs as Circuits
- Array Languages #7: X_eTaL-gpu, Array Programs on GPUs Old and New
- 256 Bytes to 640 Gigabytes: Fifty Years of Memory, in Constant Dollars
- Bucket List #4: Three Operating Systems and an ML Hardware Emulator
- Lucy 20%: Upgrading My Home AI Cluster
homelab
interpretability
introduction
language-design
- Array Languages #17: What Makes a Practical Language, IMHO
- Array Languages #16: How X_eTaL Evolved, Decision by Decision
- Array Languages #15: A Parsing Library for X_eTaL Macro Writers
- Array Languages #14: Types, Lints and Errors in X_eTaL
- Array Languages #13: Integrals and Continuous Convolution in X_eTaL
- Array Languages #12: ML Pipelines as X_eTaL Macros
- Array Languages #11: Math Equations as X_eTaL Macros
- Array Languages #10: X_eTaL on Metal and CUDA, Through Mojo
- Array Languages #9: X_eTaL and Blender, Scenes and Animation from Arrays
- Array Languages #8: X_eTaL-fpga, Array Programs as Circuits
- Array Languages #7: X_eTaL-gpu, Array Programs on GPUs Old and New
- Array Languages #6: X_eTaL-games, Rules That Are Arrays
- Array Languages #5: X_eTaL-ML, Where the Machine Learning Is
- Array Languages #4: X_eTaL-extensions, Extending the Machine
- Array Languages #3: X_eTaL-libraries, Extending the Vocabulary and the Language
- Array Languages #2: X_eTaL-demos, Programs Worth Watching
- Array Languages #1: X_eTaL, an Array Language You Can Read
- Saw #9: Espanso, Kate, ShareX, and Pluggable I2C Devices on COR24
languages
- Array Languages #17: What Makes a Practical Language, IMHO
- Array Languages #16: How X_eTaL Evolved, Decision by Decision
- Array Languages #15: A Parsing Library for X_eTaL Macro Writers
- Array Languages #14: Types, Lints and Errors in X_eTaL
- Array Languages #13: Integrals and Continuous Convolution in X_eTaL
- Array Languages #12: ML Pipelines as X_eTaL Macros
- Array Languages #11: Math Equations as X_eTaL Macros
- Array Languages #10: X_eTaL on Metal and CUDA, Through Mojo
- Array Languages #9: X_eTaL and Blender, Scenes and Animation from Arrays
- Array Languages #8: X_eTaL-fpga, Array Programs as Circuits
- Array Languages #7: X_eTaL-gpu, Array Programs on GPUs Old and New
- Array Languages #6: X_eTaL-games, Rules That Are Arrays
- Saw #14: Guards, Lints, Macros as a Test Bench, and Blender
- Array Languages #5: X_eTaL-ML, Where the Machine Learning Is
- Array Languages #4: X_eTaL-extensions, Extending the Machine
- Voxels in an Array Language: A 3D World in Dyalog APL and X_eTaL
- Array Languages #3: X_eTaL-libraries, Extending the Vocabulary and the Language
- Eigencube: Solving a Rubik's Cube with Linear Algebra, in APLSV and X_eTaL
- Array Languages #2: X_eTaL-demos, Programs Worth Watching
- Array Languages #1: X_eTaL, an Array Language You Can Read
- TBT #13: Many Decades of Learning the Next Thing --- Timelines of the Tools
- Saw #13: A Second Host in the Cloud, a Word Cloud, and Three Benches to Set Up
- Made Visible #4: Rosetta M, a Stone with Four Faces
- TBT #12: The IBM 5100's APLSV --- (B) '75, Birds by Execute, and a Machine That Learns Tic-Tac-Toe
- Saw #12: A Model That Decides, and an Index That Knows
- TBT #11: APL\360 Revisited
- AI Tools #7: A Coding Agent Small Enough to Understand
- Saw #11: Building a Tiny Mixture of Experts Microscope
- Made Visible #3: MLOS, Where Memory Means Something Else
- Made Visible #2: MesaOS, the Conventional One
- Made Visible #1: A Microkernel's Storage, Drawn
- Machine Learning #9: Teaching an Array Language to Say CNN
- Saw #10: sw-MLPL General-Purpose Features, Libraries, Extensions, Graphics, and Networking
- Saw #9: Espanso, Kate, ShareX, and Pluggable I2C Devices on COR24
- Bucket List #3: 3D Source Code, Five New Languages, and Visible Compilers
- Saw #8: Tuplet, Smalltalk-on-BASIC, proto-forth-wasm, sw-MLPL Split, and I2C on COR24
- Saw #5: Sagas, Languages, and Compiler Chains
llm
- AI Tools #6: Evaluating Local Models in a Plan-Execute-Review Loop
- AI Tools #4: Pi --- The Minimal Agent That Stays Out of the Way
- Large-Language-Monkeys: Scaling Inference, Not Models
- Five ML Concepts - #30: The Journey So Far
- Five ML Concepts - #29
- Five ML Concepts - #28
- Five ML Concepts - #27
- Five ML Concepts - #26
- Five ML Concepts - #25
- Five ML Concepts - #24
- Five ML Concepts - #23
- Five ML Concepts - #22
- Five ML Concepts - #21
- Five ML Concepts - #20
- In-Context Learning Revisited: From Mystery to Engineering
- Five ML Concepts - #19
- Five ML Concepts - #18
- Five ML Concepts - #17
- Five ML Concepts - #16
- Multi-Hop Reasoning (2/2): The Distribution Trap
- Towards Continuous LLM Learning (2): Routing Prevents Forgetting
- Five ML Concepts - #15
- Five ML Concepts - #14
- Five ML Concepts - #13
- Five ML Concepts - #12
- Five ML Concepts - #11
- RLM: Recursive Language Models for Massive Context
- Five ML Concepts - #10
- Towards Continuous LLM Learning (1): Sleepy Coder - When Fine-Tuning Fails
- Five ML Concepts - #9
- Five ML Concepts - #8
- Deepseek Papers (3/3): Engram Revisited - From Emulation to Implementation
- Five ML Concepts - #7
- Five ML Concepts - #6
- Five ML Concepts - #5
- Five ML Concepts - #4
- Five ML Concepts - #3
- Small Models (6/6): Which Small AI Fits YOUR Laptop?
- Five ML Concepts - #2
- Small Models (5/6): Max AI Per Watt
- Five ML Concepts - #1
- Small Models (4/6): This AI Has a Visible Brain
- Solving Sparse Rewards with Many Eyes
- Small Models (3/6): Planner + Doer = Genius
- Deepseek Papers (2/3): Engram - Conditional Memory for Transformers
- Multi-Hop Reasoning (1/2): Training Wheels for Small LLMs
- Small Models (2/6): AI in Your Pocket
- Deepseek Papers (1/3): mHC - Training Stability at Any Depth
- Small Models (1/6): 976 Parameters Beat Billions
machine-learning
- Array Languages #13: Integrals and Continuous Convolution in X_eTaL
- Array Languages #12: ML Pipelines as X_eTaL Macros
- Array Languages #11: Math Equations as X_eTaL Macros
- Array Languages #5: X_eTaL-ML, Where the Machine Learning Is
- Array Languages #3: X_eTaL-libraries, Extending the Vocabulary and the Language
- Array Languages #2: X_eTaL-demos, Programs Worth Watching
- Array Languages #1: X_eTaL, an Array Language You Can Read
- Saw #13: A Second Host in the Cloud, a Word Cloud, and Three Benches to Set Up
- Made Visible #4: Rosetta M, a Stone with Four Faces
- Saw #12: A Model That Decides, and an Index That Knows
- AI Tools #7: A Coding Agent Small Enough to Understand
- Saw #11: Building a Tiny Mixture of Experts Microscope
- Made Visible #3: MLOS, Where Memory Means Something Else
- Machine Learning #9: Teaching an Array Language to Say CNN
- Saw #10: sw-MLPL General-Purpose Features, Libraries, Extensions, Graphics, and Networking
- Energy-Based Learning: From Hopfield Networks to JEPA
- Saw #8: Tuplet, Smalltalk-on-BASIC, proto-forth-wasm, sw-MLPL Split, and I2C on COR24
- Large-Language-Monkeys: Scaling Inference, Not Models
- ML Frontier #05: Grokking --- Delayed Generalization
- Saw #7: Prolog, Many-Agent Isolation, Self-Hosting Assembler, and MLPL
- ML Frontier #04: Is Chain of Thought Real?
- ML Frontier #03: Structure Beats Scale --- Knowledge Graphs and Domain-Specific Superintelligence
- ML Frontier #02: In-Context Reinforcement Learning
- ML Frontier #01: Neural Collapse
- Five ML Concepts - #30: The Journey So Far
- Five ML Concepts - #29
- Five ML Concepts - #28
- How AI Learns Part 7: Designing a Continuous Learning Agent
- Five ML Concepts - #27
- How AI Learns Part 6: Toward Continuous Learning
- Five ML Concepts - #26
- How AI Learns Part 5: Context Engineering & Recursive Reasoning
- Five ML Concepts - #25
- How AI Learns Part 4: Memory-Based Learning
- Five ML Concepts - #24
- How AI Learns Part 3: Weight-Based Learning
- Five ML Concepts - #23
- How AI Learns Part 2: Catastrophic Forgetting vs Context Rot
- Five ML Concepts - #22
- Many-Eyes Learning: Intrinsic Rewards and Diversity
- How AI Learns Part 1: The Many Meanings of Learning
- Five ML Concepts - #21
- Five ML Concepts - #20
- In-Context Learning Revisited: From Mystery to Engineering
- Five ML Concepts - #19
- Five ML Concepts - #18
- Five ML Concepts - #17
- Five ML Concepts - #16
- Multi-Hop Reasoning (2/2): The Distribution Trap
- Towards Continuous LLM Learning (2): Routing Prevents Forgetting
- Five ML Concepts - #15
- Five ML Concepts - #14
- Five ML Concepts - #13
- Five ML Concepts - #12
- Neural-Net-RS: An Educational Neural Network Platform
- Cat Finder: Personal Software via Vibe Coding
- Five ML Concepts - #11
- Five ML Concepts - #10
- Towards Continuous LLM Learning (1): Sleepy Coder - When Fine-Tuning Fails
- Five ML Concepts - #9
- Five ML Concepts - #8
- Deepseek Papers (3/3): Engram Revisited - From Emulation to Implementation
- Five ML Concepts - #7
- Five ML Concepts - #6
- Five ML Concepts - #5
- Five ML Concepts - #4
- Five ML Concepts - #3
- Five ML Concepts - #2
- Five ML Concepts - #1
- Small Models (4/6): This AI Has a Visible Brain
- Solving Sparse Rewards with Many Eyes
- Small Models (3/6): Planner + Doer = Genius
- Deepseek Papers (2/3): Engram - Conditional Memory for Transformers
- Multi-Hop Reasoning (1/2): Training Wheels for Small LLMs
- Deepseek Papers (1/3): mHC - Training Stability at Any Depth
- Small Models (1/6): 976 Parameters Beat Billions
math
mcp
meta
mobile
multi-agent
operating-systems
personal
- Personal Software #11: A Campus for the Public Work
- Personal Software #10: A Sudoku That Teaches You the Strategies
- Bucket List #4: Three Operating Systems and an ML Hardware Emulator
- Bucket List #3: 3D Source Code, Five New Languages, and Visible Compilers
- Bucket List #2: A Landing Page for Software Tools
- Bucket List #1: Things I've Always Wanted to Build
process
- Array Languages #16: How X_eTaL Evolved, Decision by Decision
- Saw #15: agentrail, Correcting the Record
- Saw #14: Guards, Lints, Macros as a Test Bench, and Blender
productivity
programming
programming-history
- Eigencube: Solving a Rubik's Cube with Linear Algebra, in APLSV and X_eTaL
- 256 Bytes to 640 Gigabytes: Fifty Years of Memory, in Constant Dollars
- TBT #13: Many Decades of Learning the Next Thing --- Timelines of the Tools
- TBT #12: The IBM 5100's APLSV --- (B) '75, Birds by Execute, and a Machine That Learns Tic-Tac-Toe
- TBT #11: APL\360 Revisited
- TBT #10: Mass Compile and PL/EDIT --- 1980s Productivity Tools, Reborn on COR24 PL/SW
- TBT #9: UNIVAC Startrek, TRS-80 Adventures, and COR24 BASIC
- TBT #3: Vector Graphics Games
- TBT #2: Pipelines on OS/390
- TBT #1: My First Program Was a Horse Race
programming-languages
- Rabbit-hole #9: X_eTaL and Mojo on a Qualcomm Adreno 702
- Rabbit-hole #8: From X_eTaL to the Metal, One Kernel Down Every GPU Toolchain
- Rabbit-hole #7: The Phoenix Combinator, and Why X_eTaL Brackets Its Forks
- Rabbit-hole #6: The Sage Bird --- Y Combinators in an Eager Array Language
- Dogfooding #1: YAGNI Until You Do --- A Pascal P-Code Bump Allocator Grows a Free List
- Rabbit-hole #5: FORTH --- Retargetable Codegen and the Forward-Reference Problem
- Rabbit-hole #4: FORTH --- Dictionary Compaction and Specialized Images
- Rabbit-hole #3: FORTH --- Life After Hashing
- Rabbit-hole #2: FORTH --- FIND and the Cost of a Name
- Embedded #3: How Much of Forth Can Be Forth? A Kernel Self-Hosting Spectrum
projects
- Saw #13: A Second Host in the Cloud, a Word Cloud, and Three Benches to Set Up
- Saw #12: A Model That Decides, and an Index That Knows
- Saw #11: Building a Tiny Mixture of Experts Microscope
- Personal Software #11: A Campus for the Public Work
- Personal Software #10: A Sudoku That Teaches You the Strategies
- Bucket List #4: Three Operating Systems and an ML Hardware Emulator
- Saw #10: sw-MLPL General-Purpose Features, Libraries, Extensions, Graphics, and Networking
- Bucket List #3: 3D Source Code, Five New Languages, and Visible Compilers
- Bucket List #2: A Landing Page for Software Tools
- Bucket List #1: Things I've Always Wanted to Build
- Neural-Net-RS: An Educational Neural Network Platform
- Cat Finder: Personal Software via Vibe Coding
rabbit-hole
- Rabbit-hole #9: X_eTaL and Mojo on a Qualcomm Adreno 702
- Rabbit-hole #8: From X_eTaL to the Metal, One Kernel Down Every GPU Toolchain
- Rabbit-hole #7: The Phoenix Combinator, and Why X_eTaL Brackets Its Forks
- Rabbit-hole #6: The Sage Bird --- Y Combinators in an Eager Array Language
- Rabbit-hole #5: FORTH --- Retargetable Codegen and the Forward-Reference Problem
- Rabbit-hole #4: FORTH --- Dictionary Compaction and Specialized Images
- Rabbit-hole #3: FORTH --- Life After Hashing
- Rabbit-hole #2: FORTH --- FIND and the Cost of a Name
research
- Energy-Based Learning: From Hopfield Networks to JEPA
- Large-Language-Monkeys: Scaling Inference, Not Models
- ML Frontier #05: Grokking --- Delayed Generalization
- ML Frontier #04: Is Chain of Thought Real?
- ML Frontier #03: Structure Beats Scale --- Knowledge Graphs and Domain-Specific Superintelligence
- AI Tools #1: XSkill --- A Memory Layer for Multimodal Agents
- ML Frontier #02: In-Context Reinforcement Learning
- ML Frontier #01: Neural Collapse
- Many-Eyes Learning: Intrinsic Rewards and Diversity
- In-Context Learning Revisited: From Mystery to Engineering
- Multi-Hop Reasoning (2/2): The Distribution Trap
- Towards Continuous LLM Learning (2): Routing Prevents Forgetting
- RLM: Recursive Language Models for Massive Context
- DyTopo: Dynamic Topology for Multi-Agent AI
- Towards Continuous LLM Learning (1): Sleepy Coder - When Fine-Tuning Fails
- Deepseek Papers (3/3): Engram Revisited - From Emulation to Implementation
- Solving Sparse Rewards with Many Eyes
- Small Models (3/6): Planner + Doer = Genius
- Deepseek Papers (2/3): Engram - Conditional Memory for Transformers
- Multi-Hop Reasoning (1/2): Training Wheels for Small LLMs
- Deepseek Papers (1/3): mHC - Training Stability at Any Depth
- Small Models (1/6): 976 Parameters Beat Billions
retrocomputing
- Eigencube: Solving a Rubik's Cube with Linear Algebra, in APLSV and X_eTaL
- 256 Bytes to 640 Gigabytes: Fifty Years of Memory, in Constant Dollars
- TBT #13: Many Decades of Learning the Next Thing --- Timelines of the Tools
- TBT #12: The IBM 5100's APLSV --- (B) '75, Birds by Execute, and a Machine That Learns Tic-Tac-Toe
- TBT #11: APL\360 Revisited
- TBT #10: Mass Compile and PL/EDIT --- 1980s Productivity Tools, Reborn on COR24 PL/SW
- TBT #9: UNIVAC Startrek, TRS-80 Adventures, and COR24 BASIC
rust
- Made Visible #3: MLOS, Where Memory Means Something Else
- Personal Software #11: A Campus for the Public Work
- Made Visible #2: MesaOS, the Conventional One
- Made Visible #1: A Microkernel's Storage, Drawn
- Machine Learning #9: Teaching an Array Language to Say CNN
- Personal Software #10: A Sudoku That Teaches You the Strategies
- AI Tools #6: Evaluating Local Models in a Plan-Execute-Review Loop
- Personal Software #8: One Ring to Rule Them All --- sw-launcher's Memory Profiles, Heap Budgets, and a Working Scenario A
- AI Tools #3: sw-checklist --- Reining In AI Coding Agents With a Code-Metrics Ratchet
- AI Tools #2: AgentRail Mid-Saga --- Insert, Reorder, Reopen, Recover
- Saw #7: Prolog, Many-Agent Isolation, Self-Hosting Assembler, and MLPL
- Saw #6: Agent Coordination, Fuzzing Tests, Vendoring, and Emacs Graphics
- Saw #5: Sagas, Languages, and Compiler Chains
- Saw #4: All Together Now --- Emacs Meets the Multi-Agent Orchestra
- TBT #8: wiki-rs --- Six Wikis, One Engine, Thirty Years of History
- Saw #3: agentrail-rs --- From Walking Skeleton to Dual Memory
- Embedded #2: COR24-RS --- Learn Assembly in Your Browser
- TBT #7: reg-rs - Regression Testing from C++ to Java to Rust
- Embedded #1: BMP280 Driver --- From Prototype to Patent Proof-of-Concept
- Saw #2: reg-rs, avoid-compaction, and agentrail-rs
- Rabbit-hole #1: Poor Man's Rust-to-Unsupported-ISA Translator
- Saw #1: pjmai-rs, Rig, and langchain-rust
- pjmai-rs: Navigation History and Fuzzy Completion
- rank-wav: Ranking Audio Files by Acoustic Quality
- TBT #6: PJMAI-RS - A Shell That Knows Your Projects
- music-pipe-rs: Web Demo and Multi-Instrument Arrangements
- TBT #5: IBM 1130 System Emulator - Experience 1960s Computing
- music-pipe-rs: Unix Pipelines for MIDI Composition
- midi-cli-rs: Extending with Custom Mood Packs
- midi-cli-rs: Music Generation for AI Coding Agents
- Neural-Net-RS: An Educational Neural Network Platform
- Cat Finder: Personal Software via Vibe Coding
- RLM: Recursive Language Models for Massive Context
- DyTopo: Dynamic Topology for Multi-Agent AI
- MCP: Teaching Claude to Play (and Trash Talk)
security
self-supervised-learning
smalltalk
systems
- Made Visible #3: MLOS, Where Memory Means Something Else
- Made Visible #2: MesaOS, the Conventional One
- Made Visible #1: A Microkernel's Storage, Drawn
tbt
- TBT #13: Many Decades of Learning the Next Thing --- Timelines of the Tools
- TBT #12: The IBM 5100's APLSV --- (B) '75, Birds by Execute, and a Machine That Learns Tic-Tac-Toe
- TBT #11: APL\360 Revisited
- TBT #10: Mass Compile and PL/EDIT --- 1980s Productivity Tools, Reborn on COR24 PL/SW
- TBT #9: UNIVAC Startrek, TRS-80 Adventures, and COR24 BASIC
- TBT #8: wiki-rs --- Six Wikis, One Engine, Thirty Years of History
- TBT #7: reg-rs - Regression Testing from C++ to Java to Rust
- TBT #6: PJMAI-RS - A Shell That Knows Your Projects
- TBT #5: IBM 1130 System Emulator - Experience 1960s Computing
- TBT #4: ToonTalk - Teaching Robots to Program
- TBT #3: Vector Graphics Games
- TBT #2: Pipelines on OS/390
- TBT #1: My First Program Was a Horse Race
testing
- Saw #6: Agent Coordination, Fuzzing Tests, Vendoring, and Emacs Graphics
- TBT #7: reg-rs - Regression Testing from C++ to Java to Rust
tools
- Array Languages #15: A Parsing Library for X_eTaL Macro Writers
- Array Languages #14: Types, Lints and Errors in X_eTaL
- Array Languages #13: Integrals and Continuous Convolution in X_eTaL
- Saw #15: agentrail, Correcting the Record
- Array Languages #12: ML Pipelines as X_eTaL Macros
- Array Languages #11: Math Equations as X_eTaL Macros
- Array Languages #10: X_eTaL on Metal and CUDA, Through Mojo
- Array Languages #9: X_eTaL and Blender, Scenes and Animation from Arrays
- Array Languages #6: X_eTaL-games, Rules That Are Arrays
- Saw #14: Guards, Lints, Macros as a Test Bench, and Blender
- Array Languages #5: X_eTaL-ML, Where the Machine Learning Is
- Array Languages #4: X_eTaL-extensions, Extending the Machine
- Array Languages #3: X_eTaL-libraries, Extending the Vocabulary and the Language
- Array Languages #2: X_eTaL-demos, Programs Worth Watching
- Array Languages #1: X_eTaL, an Array Language You Can Read
- TBT #13: Many Decades of Learning the Next Thing --- Timelines of the Tools
- Saw #13: A Second Host in the Cloud, a Word Cloud, and Three Benches to Set Up
- Saw #12: A Model That Decides, and an Index That Knows
- AI Tools #7: A Coding Agent Small Enough to Understand
- AI Tools #6: Evaluating Local Models in a Plan-Execute-Review Loop
- AI Tools #4: Pi --- The Minimal Agent That Stays Out of the Way
- AI Tools #5: nono --- Sandboxing Pi Without Breaking the Loop
- Saw #9: Espanso, Kate, ShareX, and Pluggable I2C Devices on COR24
- AI Tools #1: XSkill --- A Memory Layer for Multimodal Agents
- Saw #2: reg-rs, avoid-compaction, and agentrail-rs
- Saw #1: pjmai-rs, Rig, and langchain-rust
- music-pipe-rs: Web Demo and Multi-Instrument Arrangements
- music-pipe-rs: Unix Pipelines for MIDI Composition
- midi-cli-rs: Extending with Custom Mood Packs
- midi-cli-rs: Music Generation for AI Coding Agents
vibe-coding
- music-pipe-rs: Web Demo and Multi-Instrument Arrangements
- TBT #5: IBM 1130 System Emulator - Experience 1960s Computing
- Many-Eyes Learning: Intrinsic Rewards and Diversity
- music-pipe-rs: Unix Pipelines for MIDI Composition
- midi-cli-rs: Extending with Custom Mood Packs
- midi-cli-rs: Music Generation for AI Coding Agents
- TBT #4: ToonTalk - Teaching Robots to Program
- Neural-Net-RS: An Educational Neural Network Platform
- Cat Finder: Personal Software via Vibe Coding
- TBT #3: Vector Graphics Games