Zachary Proser

HOVER PREVIEW · CLICK TRAVEL · T TOUR · ← → BROWSE · ENTER READ

SHT 00 · The index · 185 essays plotted

The mind is a fire,to be ignited

Research, blueprints, workshops, working systems.

The sky above plots all 185 essays. Explore the full constellation on a larger screen, or browse every essay below.

✱The Modern Coding letter5,000+ engineers
Dispatches from the edge of applied AI.

What lands in your inbox: what I’m building, what I’m learning, the tools I’d actually pay for. Names named, vendors graded, evals included.

Unsubscribe in one click · No spam, ever
Read all my writing →·Workshops & AI enablement →
ZRead essay →
PLATE 00 — THE CORPUS · ALWAYS-NIGHT CANVAS, MOUNTED185 STARS
ZP
Title
INDEX OF WORK — ZACKPROSER.COM
Drawn
ZP
Scale
1:1
Rev
G
Date
NOV 2026
Sht
00 / 07
Who's writing
Portrait of Zachary Proser
Plate I · MMXXVI
I do this work because I love it — and everything I learn, I give away.

Hi, I'm Zachary.

I work on the Applied AI team at WorkOS, shipping internal and customer-facing systems, automations, and apps that cut toil — and bringing the rest of the org along with workshops on AI fundamentals, training, office hours, and support.

I'm also an independent AI researcher working on local inference, inference engineering, and cost optimization. Through my consultancy, Mind on Fire, I do custom workshops, software and systems development, and bespoke training. As a Hone instructor, I regularly deliver courses and interactive workshops on AI enablement, organizational AI adoption, and balancing productivity with rest. For the last two years I've taught engineers everything from inference to agentic development, loops, scheduling, and skills at scale at the AI Engineer conferences in London, New York, and San Francisco.

More about me →·Say hello →
Currently
Applied AI · WorkOS
Teaching
AI instructor · Hone
Consultancy
Mind on Fire
Researching
Local inference, inference engineering, cost optimization
18
Blueprint drawings
Schematics · equations · live demos, § 01 →
60
Runs, one coding-agent benchmark
Wilson 95% CI · inspect every run →
6,000+
Newsletter subscribers
The Modern Coding Letter · weekly →
12.4M
Search impressions / yr
Search Console · trailing 12 mo →
ZP · SHT 01 — The Blueprint Folio · Dossier18 plates · PL-001 — PL-020

Applied AI, drawn to scale.

I love to research, experiment and publish my learnings as interactive experiences for everyone to benefit from.

Open the folio →
The Transformer blueprint platePL-001ATTENTIONThe TransformerEvery frontier model runs the same circuit. This working drawing traces it end to end: how seq2seq attention became the transformer, how decoder-only models generate, and how tokens, attention heads, position, residual blocks, sampling, and KV caching fit together.Schematic · Worked math · 15 MINRead the plate →The Embedding Space blueprint platePL-002EMBEDDINGSThe Embedding SpaceFrom the distributional hypothesis and word2vec to contextual sentence encoders: how embeddings are trained, compared, indexed with HNSW, migrated, and evaluated in production.Schematic · Citations · 15 MINOpen instrument ▸The RAG Pipeline blueprint platePL-003RETRIEVALThe RAG PipelineFrom open-domain question answering to production RAG: ingestion, chunking, BM25 and dense retrieval, reranking, context assembly, citations, evaluation, and failure tracing.Schematic · Citations · 15 MINOpen instrument ▸The Tokenizer blueprint platePL-004TOKENIZATIONThe TokenizerHow tokenization works: BPE tokenizer training, ordered merges, byte fallback, token IDs, costs, multilingual tradeoffs, and why LLMs use tokens.Schematic · Worked math · 15 MINOpen instrument ▸The Workshop blueprint platePL-005AI ENABLEMENTThe WorkshopA complete working drawing for an AI training workshop: how to design, build, deliver, and follow through on an AI workshop for engineers or corporate AI training that works.Schematic · Field notes · 15 MINRead the plate →The Model Sheet blueprint platePL-006MODEL SELECTIONThe Model SheetA model's name is a datasheet most people never learn to read. This drawing goes field by field — parameters, dense versus MoE, training stage, distillation, abliteration and merges, quantization (GGUF, k-quants, imatrix, GPTQ/AWQ) — then does the memory math that decides what runs locally, sources and vets a model on Hugging Face, and tests it on your own task instead of trusting a benchmark.Schematic · Worked math · 20 MINRead the plate →The Attention Head blueprint platePL-010INTERPRETABILITYThe Attention HeadWhat is an attention head, and how would you prove one does what you think it does? A from-zero walk through a single head in GPT-2 small: what it reads, what it writes, the copying behavior you can watch in its attention map, and why that picture is the weakest evidence there is — with logit attribution and ablation as the stronger tests.Schematic · Worked math · 15 MINRead the plate →The Inference Engine blueprint platePL-011SERVINGThe Inference EngineA worked guide to LLM inference memory, KV-cache paging, continuous batching, prefill, decode, quantization, and latency-capacity planning.Schematic · Worked math · 15 MINRead the plate →The Guard blueprint platePL-012AGENT SECURITYThe GuardA trust-boundary threat model for prompt injection in tool-using agents, with capability scoping, external authorization, and testable security metrics.Schematic · Citations · 15 MINRead the plate →The Diffusion Model blueprint platePL-013GENERATIVE VISIONThe Diffusion ModelA coordinate-consistent visual guide to diffusion and flow: forward noise, learned denoising, guidance, latent compression, DiT, sampling, and flow matching.Schematic · Citations · 15 MINRead the plate →The Benchmark blueprint platePL-014EVALUATIONThe BenchmarkA benchmark score is manufactured, and every step leaves a trace. Trace an AI benchmark number back through sampled tasks, prompts, scorers, aggregation, uncertainty, and contamination to the decision it can actually support.Schematic · Statistics · 15 MINRead the plate →CTX blueprint platePL-015PERSONAL CONTEXTCTXEvery agent I use — terminal, browser, phone, across Anthropic and OpenAI — started every conversation knowing nothing about me. So I built one shared memory they all plug into: a technical blueprint for a provider-agnostic personal context engine — one Postgres table, a hybrid retriever with an LLM reranker, and a scope wall enforced by the database, not by prompts.Schematic · Field notes · 30 MINRead the plate →The Voice Box blueprint platePL-016LOCAL SPEECHThe Voice BoxOpen-weight speech synthesis got good enough to replace a paid API. This is the working drawing for a local text-to-speech studio built on Fish Audio S2 Pro and MLX — every measurement, both bugs, and the discovery that the model ships with no voices at all.Schematic · Field notes · 15 MINRead the plate →The Cost Curve blueprint platePL-017MODEL ECONOMICSThe Cost CurveDeepSeek's retrained V4 Flash 0731 moved the model pricing baseline for everyone. This drawing works the full token-dollar ledger: MoE bandwidth arithmetic, quantization, the nine agent benchmarks where a budget model beat the flagship, cache-hit economics, verbosity-adjusted task cost, the price war forcing 80% cuts on the frontier, and the structural trap forcing labs to race their own depreciation curve.Schematic · Worked math · 25 MINRead the plate →The Price Floor blueprint platePL-018INFERENCE COSTThe Price FloorWhat does it actually cost to run a frontier-class open model? I measured DeepSeek V4 Flash 0731 across every layer in August 2026: why the same weights run 5x faster in a different inference engine, prefill versus decode, why speculative-decoding gains don't transfer between stacks, measured cloud throughput against the listed figures, the prompt-cache price that dominates agent workloads, the sold-out market for fast inference, and the monthly crossover where a home Blackwell rig beats the API.Schematic · Worked math · 15 MINRead the plate →The Eval Harness blueprint platePL-019AGENT FACTORYThe Eval HarnessI built pt, an attention-first agentic terminal in Go: durable work contracts, parallel agent lanes with byte-verified delivery, workers on rented orbs. The factory merged nearly 400 PRs in two weeks, and I still couldn't drive my workday through it. This is the blueprint and the retro: what got built, what failed to get traction, the two ctx-backed capabilities that stuck, and the new direction: ctx as a complete, verifiable graph of work that any harness or future model can drive.Schematic · Statistics · 26 MINOpen instrument ▸The Eval Harness blueprint platePL-019EVALUATIONThe Eval HarnessA complete record of 60 coding-agent runs across an 80.76 GiB local DeepSeek quant, hosted DeepSeek, and hosted Claude: task design, hidden scorers, repair passes, provider recovery, exact cost, uncertainty, and workload fit.Schematic · Statistics · 20 MINOpen instrument ▸Designing AI Evaluations blueprint platePL-020AI EVALUATIONDesigning AI EvaluationsA generic blueprint for designing AI evaluations: define the decision, version the system, build representative task sets, choose valid scorers, measure uncertainty, analyze failures, and connect offline tests to production evidence.Schematic · Statistics · 25 MINRead the plate →
ZP · SHT 01A — Exploded Views · Explainers5 plates · PL-EX-001 — PL-EX-005

The whole machine, pulled apart.

Exploded-view explainers that take one everyday act — asking an AI a question, loading a web page, calling a tool, pushing a commit, signing in with a passkey — and separate every layer so you can watch the request travel through each one.

All demos →
The Inference Stack exploded-view platePL-EX-001The Inference StackHow an AI response gets madeFollow one request through the software and hardware that runs a language model, then watch each generated token return to the chat box.Exploded view · InteractiveOpen instrument ▸The Fetch exploded-view platePL-EX-002The FetchWhat happens when you fetch a URL?Follow one browser navigation through URL parsing, DNS, the local network, internet routing, encryption, edge delivery, the origin server, and the rendering pipeline.Exploded view · InteractiveOpen instrument ▸The Tool Call exploded-view platePL-EX-003The Tool CallWhat happens when an agent calls a tool?Follow one function call from a model’s tool_use block through schema checks, permission, credentials, transport, and execution, then watch the observation return.Exploded view · InteractiveOpen instrument ▸The Push exploded-view platePL-EX-004The PushWhat happens when you git push?Follow one commit from git push origin HEAD through refspecs, a thin pack, HTTPS, receive-pack, and the ref update, then watch the status line come back.Exploded view · InteractiveOpen instrument ▸The Passkey exploded-view platePL-EX-005The PasskeyWhat happens when you sign in with a passkey?Follow one passkey login from the username field through a WebAuthn challenge, the authenticator, an assertion signature, and the session cookie that comes back.Exploded view · InteractiveOpen instrument ▸
ZP · SHT 02 — The Bench · WorkshopsFrom $15K · Q4 open
§ 02 · Workshops · From $15kQ4 open

Taught from the bench, not the slide deck.

Book a workshop →Engagement shapes →
Zachary Proser and Nick Nisi presenting Skills at Scale to a packed room at AI Engineering LondonZachary Proser presenting at the Claude Cowork workshopThe workshop room at Moscone West for Lifestyles of the AI-NativeAttendees coding along during the Claude Cowork workshop

For the past two years, I've been on the road teaching AI to engineers in hands-on workshops at the AI Engineer conferences in London, New York, and San Francisco.

The workshop material comes straight from the research above — real cost curves, evals your team can rerun. We embed inside your pods, ½ to 2 days, and leave you with working workflows and named champions.

⟡ Built with Anthropic⟡ 800 registrations, last public run⟡ “The gold standard for AI Engineer content” — swyx⟡ Taught at AIE World's Fair SF · London · WorkOS × Anthropic

§ 03Publications

Full list →
arXiv · 2510.25819 · Oct 2025
Identity Management for Agentic AI
With Tobin South, Alex Pentland (MIT), and others.
ChapterRAG Evaluation — official contribution to Pinecone’s Vector Databases in Production book series.
50+ articlesTechnical writing at WorkOS, Pinecone, Cloudflare, Gruntwork — agents, RAG, auth, vector DBs, IaC.
180+ essayszackproser.com/blog, plus The New Stack, Prefect.io, and the Pinecone Learning Center.
§ 04

On stage

All talks →
Untethered ProductivityFeatured · AI Engineering London · April 2026Untethered ProductivityWatch · 32 min
Modern AI and SecurityDevSecCon 2025 · KeynoteModern AI and SecurityClaude Cowork GTM WorkshopWorkOS × AnthropicClaude Cowork GTM WorkshopSkills at ScaleAIE London · WorkshopSkills at Scale
AIE World's Fair — SFAI Engineering LondonDevSecCona16zWorkOS × AnthropicHone instructor

Teaching AI enablement through Hone

I teach professionals and engineering teams how to use AI in daily work, recognize common adoption pitfalls, and build productive routines that leave room for rest. Recent classes cover planning your day with AI, sustainable agent workflows, and engineering in regulated environments.

Explore my Hone courses & workshops →
§ 05

Career

14+ Years Engineering Experience
WorkOSApplied AI2025–nowAI workflows, internal tooling, developer education. Created the Claude Cowork workshop with Anthropic.
WorkOSDeveloper Education2024–25Built and led the developer-education surface at WorkOS — documentation, workshops, and video.
PineconeStaff DevRel2023–24RAG pipelines, AWS reference architectures, dev education during the vector DB explosion.
GruntworkTech Lead2020–23Built git-xargs (1K★), contributed to cloud-nuke (3.1K★). Multi-account AWS, production IaC.
CloudflareSenior SWE2017–20Joined at ~100 engineers. Built cf-terraforming (1.3K★). APIs, edge infra, dev tools at scale.

Regularly advising hedge funds and PE firms on AI infrastructure via AlphaSights, Guidepoint, Tegus, GLG, Coleman.

§ 05 · Measured reach — trailing 12 moCareer record above
12.4MSearch impressions / yr
52.5KSearch clicks · avg pos 7.9
137KVisitors / yr · +21% YoY
179KPage views / yr · +18%
6,000+Newsletter subscribers
0Search Console + Plausible · trailing 12 mo12.4M
🔥 Mind on Fire — independent AI consultancy
Local inference economics, eval methodology, and systems from silicon to prompt.
mindonfire.net →
Drawing set · Issued NOV 2026Sht 00 of 07Next sheet · 01 WRITING →