DEV Community

Synthetic Scars Series' Articles

Back to Randal L. Schwartz's Series
Why AI Keeps Making the Same Coding Mistakes—And How Teaching It Pain Gives It Wisdom
Cover image for Why AI Keeps Making the Same Coding Mistakes—And How Teaching It Pain Gives It Wisdom

Devs compare AI to a 3 AM incident veteran
randalschwartz profile

Why AI Keeps Making the Same Coding Mistakes—And How Teaching It Pain Gives It Wisdom

#ai #programming #architecture #productivity
27
Comments 11
10 min read
Why AI Coding Agents Crash at 3 AM: The Happy-Path Mirage & The Forced Continuity Defect
Cover image for Why AI Coding Agents Crash at 3 AM: The Happy-Path Mirage & The Forced Continuity Defect

Real agent insights validate silent failures
randalschwartz profile

Why AI Coding Agents Crash at 3 AM: The Happy-Path Mirage & The Forced Continuity Defect

#ai #programming #architecture #productivity
15
Picked as gem Comments 16
11 min read
The Physics of Socratic Prompting: Somatic Recoil, Chess Alpha-Beta, & The NLP Meta-Model
Cover image for The Physics of Socratic Prompting: Somatic Recoil, Chess Alpha-Beta, & The NLP Meta-Model

Uses a Google Maps analogy for bad RLHF routing
randalschwartz profile

The Physics of Socratic Prompting: Somatic Recoil, Chess Alpha-Beta, & The NLP Meta-Model

#ai #programming #architecture #machinelearning
15
Picked as gem Comments 7
17 min read
Nudging with Questions: Why Telling Your AI What to Fix Triggers an Apology Death Spiral (And How to Advance Juniors)
Cover image for Nudging with Questions: Why Telling Your AI What to Fix Triggers an Apology Death Spiral (And How to Advance Juniors)

Apology loops waste context window tokens
randalschwartz profile

Nudging with Questions: Why Telling Your AI What to Fix Triggers an Apology Death Spiral (And How to Advance Juniors)

#ai #programming #architecture #productivity
17
Comments 14
16 min read
Stuart Feldman Was Right in 1976: Why Your AI Agent Needs a Makefile, Not a 20-Step Prompt
Cover image for Stuart Feldman Was Right in 1976: Why Your AI Agent Needs a Makefile, Not a 20-Step Prompt
randalschwartz profile

Stuart Feldman Was Right in 1976: Why Your AI Agent Needs a Makefile, Not a 20-Step Prompt

#ai #programming #architecture #productivity
15
Picked as gem Comments 22
11 min read
Surviving the 200k-Token Lobotomy: How Unix init.d and 'Memento' Made My AI Coding Agent Immune to Context Compaction
Cover image for Surviving the 200k-Token Lobotomy: How Unix init.d and 'Memento' Made My AI Coding Agent Immune to Context Compaction

External state files prevent token loss
randalschwartz profile

Surviving the 200k-Token Lobotomy: How Unix init.d and 'Memento' Made My AI Coding Agent Immune to Context Compaction

#ai #programming #architecture #unix
8
Comments 25
15 min read
What LLMs 'Know' That You Don't Know They Know: Stop Inventing Prompt DSLs and Ride 50 Years of Unix Pre-Training Gravity
Cover image for What LLMs 'Know' That You Don't Know They Know: Stop Inventing Prompt DSLs and Ride 50 Years of Unix Pre-Training Gravity

Real utility execution beats text prompts
randalschwartz profile

What LLMs 'Know' That You Don't Know They Know: Stop Inventing Prompt DSLs and Ride 50 Years of Unix Pre-Training Gravity

#ai #programming #unix #architecture
32
Comments 48
22 min read
If You Never Do Anything Else to Fix Your Prompts, Do This One Thing
Cover image for If You Never Do Anything Else to Fix Your Prompts, Do This One Thing

Links 1986 linguistic theory to LLM hallucinations
randalschwartz profile

If You Never Do Anything Else to Fix Your Prompts, Do This One Thing

#ai #programming #architecture #productivity
15
Comments 7
17 min read
Requests, Promises, Assertions, and Declarations: Rewriting Our AI Agent Workflow as a 1986 'Committed Speech' Machine
Cover image for Requests, Promises, Assertions, and Declarations: Rewriting Our AI Agent Workflow as a 1986 'Committed Speech' Machine

Portland 1986 roots of modern agent workflows
randalschwartz profile

Requests, Promises, Assertions, and Declarations: Rewriting Our AI Agent Workflow as a 1986 'Committed Speech' Machine

#ai #programming #architecture #unix
27
Comments 43
26 min read
926 Steps, Two Lobotomies, and Zero Nudges: Flight-Testing the Commitment Machine Across Gemini, Claude, and GPT-OSS
Cover image for 926 Steps, Two Lobotomies, and Zero Nudges: Flight-Testing the Commitment Machine Across Gemini, Claude, and GPT-OSS

Unix flat-files beat memory compaction
randalschwartz profile

926 Steps, Two Lobotomies, and Zero Nudges: Flight-Testing the Commitment Machine Across Gemini, Claude, and GPT-OSS

#ai #programming #architecture #unix
34
Comments 52
17 min read
The Best Prompt Rule Is the One You Delete: Moving AI Guardrails Out of Markdown and Into Git Locks, Linux insmod, and 15ms Compiler Auto-Fixes
Cover image for The Best Prompt Rule Is the One You Delete: Moving AI Guardrails Out of Markdown and Into Git Locks, Linux insmod, and 15ms Compiler Auto-Fixes
randalschwartz profile

The Best Prompt Rule Is the One You Delete: Moving AI Guardrails Out of Markdown and Into Git Locks, Linux insmod, and 15ms Compiler Auto-Fixes

#ai #programming #architecture #compilers
7
Comments 64
29 min read
Giving AI Pain: The Architecture of Synthetic Scars & The Rapid-Regret Miner
Cover image for Giving AI Pain: The Architecture of Synthetic Scars & The Rapid-Regret Miner

Connects token burn loops to junior dev habits
randalschwartz profile

Giving AI Pain: The Architecture of Synthetic Scars & The Rapid-Regret Miner

#ai #programming #architecture #unix
25
Picked as gem Comments 44
29 min read
All That Glisters Is Softmax: Shiny-Object Interns, Airplane-Magazine Execs, and Why AI Labs Can Only Bake In Half the Scars
Cover image for All That Glisters Is Softmax: Shiny-Object Interns, Airplane-Magazine Execs, and Why AI Labs Can Only Bake In Half the Scars

Explores gradient cancellation limits
randalschwartz profile

All That Glisters Is Softmax: Shiny-Object Interns, Airplane-Magazine Execs, and Why AI Labs Can Only Bake In Half the Scars

#ai #programming #architecture #machinelearning
17
Comments 28
29 min read
The Postcard That Ate Six Grammars: 1980 Smalltalk-80, 1976 Make, TDD, and the Four Attractors of Agentic Coding
Cover image for The Postcard That Ate Six Grammars: 1980 Smalltalk-80, 1976 Make, TDD, and the Four Attractors of Agentic Coding

Randal Schwartz weighs in on LLM control surfaces
randalschwartz profile

The Postcard That Ate Six Grammars: 1980 Smalltalk-80, 1976 Make, TDD, and the Four Attractors of Agentic Coding

#ai #architecture #smalltalk #programming
10
Comments 9
43 min read
Crows, Moths, and Makefiles: How 300 Megabytes of Crow DNA Explains Why a 20-Kilobyte Scar File Outsmarts a Terabyte LLM
Cover image for Crows, Moths, and Makefiles: How 300 Megabytes of Crow DNA Explains Why a 20-Kilobyte Scar File Outsmarts a Terabyte LLM

Comments debate bare-metal local inference
randalschwartz profile

Crows, Moths, and Makefiles: How 300 Megabytes of Crow DNA Explains Why a 20-Kilobyte Scar File Outsmarts a Terabyte LLM

#ai #programming #neuroscience #architecture
20
Comments 12
13 min read
Being, Doing, Having: Why 50 Years of Senior Craft Lives in Context, Not Code
Cover image for Being, Doing, Having: Why 50 Years of Senior Craft Lives in Context, Not Code

Compares LLM prompting to stand-up comedy
randalschwartz profile

Being, Doing, Having: Why 50 Years of Senior Craft Lives in Context, Not Code

#ai #programming #architecture #productivity
18
Picked as gem Comments 17
29 min read