DEV Community

The AI Commit: daily AI engineering lab Series' Articles

Back to Mayur Agnani's Series
How to Make Your Agent Check the Records It Changed Before It Says Done
Cover image for How to Make Your Agent Check the Records It Changed Before It Says Done

How to Make Your Agent Check the Records It Changed Before It Says Done

Comments
3 min read
How to Stop a Coding Agent and an AI Reviewer From Looping on One Pull Request
Cover image for How to Stop a Coding Agent and an AI Reviewer From Looping on One Pull Request

How to Stop a Coding Agent and an AI Reviewer From Looping on One Pull Request

Comments
3 min read
How Sorting Your Prompts Speeds Up a Mixture-of-Experts Batch Job
Cover image for How Sorting Your Prompts Speeds Up a Mixture-of-Experts Batch Job

How Sorting Your Prompts Speeds Up a Mixture-of-Experts Batch Job

Comments
3 min read
Why Speculative Decoding Slows Hybrid Mamba Models at Large Batch Sizes
Cover image for Why Speculative Decoding Slows Hybrid Mamba Models at Large Batch Sizes

Why Speculative Decoding Slows Hybrid Mamba Models at Large Batch Sizes

Comments
3 min read
Why Split Prefill and Decode GPUs Lose Time in the Hand-Off, Not the Network
Cover image for Why Split Prefill and Decode GPUs Lose Time in the Hand-Off, Not the Network

Why Split Prefill and Decode GPUs Lose Time in the Hand-Off, Not the Network

Comments
3 min read
Why Forced JSON Output Is Always Valid and Sometimes Wrong
Cover image for Why Forced JSON Output Is Always Valid and Sometimes Wrong

Why Forced JSON Output Is Always Valid and Sometimes Wrong

Comments
3 min read
How to Compare Model Prices When Each Model Counts Tokens Differently
Cover image for How to Compare Model Prices When Each Model Counts Tokens Differently

How to Compare Model Prices When Each Model Counts Tokens Differently

Comments
3 min read
How to Correct an LLM Judge's Pass Rate With 100 Human Labels
Cover image for How to Correct an LLM Judge's Pass Rate With 100 Human Labels

How to Correct an LLM Judge's Pass Rate With 100 Human Labels

Comments
3 min read
How Claude Code and Codex Decide Which Skill Descriptions the Model Reads
Cover image for How Claude Code and Codex Decide Which Skill Descriptions the Model Reads

How Claude Code and Codex Decide Which Skill Descriptions the Model Reads

Comments
3 min read
How to Find the Agent Actions a Prompt Injection Can Control
Cover image for How to Find the Agent Actions a Prompt Injection Can Control

How to Find the Agent Actions a Prompt Injection Can Control

1
Comments
3 min read
How Many Pull Requests a Merge Queue Can Land in an Hour
Cover image for How Many Pull Requests a Merge Queue Can Land in an Hour

How Many Pull Requests a Merge Queue Can Land in an Hour

Comments
3 min read
How to Tell Whether a Second Model Reviewing Your Agent's Work Pays for Itself
Cover image for How to Tell Whether a Second Model Reviewing Your Agent's Work Pays for Itself

How to Tell Whether a Second Model Reviewing Your Agent's Work Pays for Itself

Comments
3 min read
How to Tell What Claude Code and Codex Forget When They Compact a Session
Cover image for How to Tell What Claude Code and Codex Forget When They Compact a Session

How to Tell What Claude Code and Codex Forget When They Compact a Session

Comments
3 min read
How to Stop a Long LLM Answer From Freezing When the Server Runs Out of Memory
Cover image for How to Stop a Long LLM Answer From Freezing When the Server Runs Out of Memory

How to Stop a Long LLM Answer From Freezing When the Server Runs Out of Memory

Comments
3 min read
How to Fix Jargon in Speech-to-Text Without Adding Words Nobody Said
Cover image for How to Fix Jargon in Speech-to-Text Without Adding Words Nobody Said

How to Fix Jargon in Speech-to-Text Without Adding Words Nobody Said

1
Comments 1
3 min read
How to Choose How Many Search Results a Reranker Should Rescore
Cover image for How to Choose How Many Search Results a Reranker Should Rescore

How to Choose How Many Search Results a Reranker Should Rescore

Comments
3 min read
How to Send Claude Only the New Messages in a Long Agent Loop
Cover image for How to Send Claude Only the New Messages in a Long Agent Loop

How to Send Claude Only the New Messages in a Long Agent Loop

Comments
3 min read
How to Stop a Model Picking the Wrong Fact From a Long Prompt
Cover image for How to Stop a Model Picking the Wrong Fact From a Long Prompt

How to Stop a Model Picking the Wrong Fact From a Long Prompt

Comments
3 min read
How to Predict Each User's Token Speed on a Shared GPU
Cover image for How to Predict Each User's Token Speed on a Shared GPU

How to Predict Each User's Token Speed on a Shared GPU

Comments
3 min read
How a Coding Agent Finds Code: Grep Versus an Embedding Index
Cover image for How a Coding Agent Finds Code: Grep Versus an Embedding Index

How a Coding Agent Finds Code: Grep Versus an Embedding Index

Comments 2
3 min read