Your banned-word list expired in March 2024
Naming a tell is what kills it, so any list of AI words starts dying the day it ships. Underneath sits the failure no list touches: the draft commits to nothing. I built an editing skill for that one.
Alex Chernysh — Writing
Operator notes on grounded AI systems, evals, retrieval, and integration-heavy delivery.
Latest posts
RightLayout: Shipping a Mac AI Tool, Then Letting Go
Why I trained a small CoreML model from scratch for a Mac keyboard-layout corrector, used it for months, and then open-sourced it instead of scaling it.
Forecasting Without Prophecy: a plain-text discipline
Why I leave the future to astrology and reach for reference classes, premortems, and calibration logs instead. Disciplined uncertainty in plain text.
Need a job? Sip your drink. We'll look for you.
Why I stopped applying the LinkedIn way and built a quiet service that does the grind for me, and now for you. Drop a résumé once, get a ranked daily shortlist with a one-line pitch hint per role.
I Ran 12 AI Agents for 47 Hours. Here's What Survived.
Open-source deterministic orchestrator for parallel CLI coding agents. Runs Claude Code, Codex CLI, Gemini CLI in parallel: zero coordination tokens, 40+ adapters, janitor verification, git worktree isolation.
Working Under Repeated Alarms
A short note from Israel on what repeated alarms do to attention, engineering judgment, and team habits, and which working practices make interruption easier to absorb.
How to Build Legal Answering Systems That Can Be Trusted
A practical blueprint for legal QA, shaped in part by work around the Agentic RAG Legal Challenge: document identity, hybrid retrieval, structured answers, page-level grounding, telemetry, and evals.
LLM Product Safety Without Theater
A practical guide to LLM product safety: prompt injection, excessive agency, unsafe outputs, evals, and sober boundaries.
Interface Design for Serious Products
A practical memo on calm authority, visible product care, restrained motion, and why trustworthy interfaces feel expensive.
Getting AI-Assisted Development to Green Without Breaking the Code
Repair loops, small diffs, test trust, and how to get CI back to green without trashing the codebase.
Building Agentic AI Systems That Hold Up
Practical guidance on tool contracts, context engineering, evals, approvals, and telemetry.
Which Query Transformation Techniques Actually Help RAG?
HyDE, query rewrite, decomposition, step-back prompting, and fusion for RAG: which query transformation technique fixes which retrieval failure, and when the extra latency pays off.
Preventing Hallucinations in LLM Systems
How to reduce hallucinations in LLM systems with better retrieval, abstention, verification, evals, and guardrails.
Most RAG Failures Start in the Documents
Chunking, titles, metadata, parent-child structure, reranking, and corpus QA for RAG systems.
Spec-Driven Development: the workflow I actually use
How I use a lightweight spec-driven workflow in real projects, what SDDRush automates, and where Kotef fits if you want a stronger agent layer.
How to Run LLM Evals in Production
LLM evals for continuous delivery: turn production failures into automated tests, grade traces with task-specific graders, and block bad releases with eval-driven gates.
Prompt Engineering: From Phrasing to Policy
Prompt design now means response formats, examples, tools, and eval loops, not incantations.
BI Storytelling That Actually Moves Decisions
How to make BI pages support decisions through narrative, visual hierarchy, and trust.
SYNAPSE: adaptive-control software engineering, prototyped
SYNAPSE was a 2025 framework for AI agents that adapt their own success criteria via MCDM. The deterministic-control-plane idea later shipped as Bernstein.