mikias\abera

Mikias Abera · Senior product engineer · Toronto

Complex systems, built into clear products.

I've spent 10+ years shipping React and TypeScript products, building from zero to production, and leading a company through acquisition. Now I build AI and data systems that verify their outputs.

Follow the workread the essays →

sourcecheck · spot-check, replayed

$ verify --claim c_28812 --doi 10.1038/s41586-…

resolving doi ............... ok

fetching abstract ........... ok · 214ms

matching claim → source ..... aligned

direction of effect ......... consistent

VERDICT ▸ SUPPORTED

$ verify --claim c_28813 --doi 10.1016/j.cortex…

resolving doi ............... ok

source states no magnitude .. cannot confirm

VERDICT ▸ NOT VERIFIABLE — abstained

$ ▌

Real behavior — 4 of 5 citations failed the first spot-check.
The gate is why 0 shipped.

10+ years

shipping production software

Founder → acquisition

end-to-end ownership

React · TypeScript · AI

product systems

AI systems where being wrong is expensive

The orange mark in each diagram is the failure the system exists to catch

NoeticMap

65,074 papers triaged to 2,233 fully extracted and 10,063 insights — anything unverifiable quarantined, not guessed at.

0 UNVERIFIED PUBLISHED

sourcecheck

A citation gate that caught 4 of 5 fabricated references on its first spot-check.

OPEN SOURCE · POWERS THE CITATION BOARD

Enacted

Clause-level diffs of Ontario law, computed from source text rather than summarized into existence.

EVERY DIFF TRACEABLE

PermitCheck

Toronto permit pre-checks where every verdict cites the exact bylaw line it rests on.

RULES DECIDE, MODEL DRAFTS

all projects →

LIVE

Running measurements, public audits

registry lookups · updated monthly

THE BOARD

Citation Integrity

489 citations checked · latest run 2026-07-03

Which frontier models fabricate citations, measured by registry lookup, never by another model's opinion.

See the board

THE PLAYGROUND

Watch it verify

Live resolution · nothing mocked

Pick a model and a question, then watch every citation resolve or get flagged against the registries.

Open the playground

IN THE WILD

The incident record

1,737 incidents tracked · court decisions · reports · policy docs

Where fabricated citations actually shipped: courts, consulting reports, national policy. Growing daily.

Browse incidents

WRITING

Build logs and failure reports

From production · one essay a week

verification2026-07-01

Same words, opposite claim

I built a checker that verifies AI claims against the real papers they cite, then paid two frontier models to break it. They did, twice. Both breaks were the same lesson: grounding checkers fail in the direction of agreement, and the fix is to accept less, not to get smarter.

fundamentals2026-07-01

Red-teaming my own pipelines: what a regulated buyer asks first

I walked my own systems the way a hostile reviewer would, from the prompt injection surface to the PII boundaries a regulated buyer asks about first, and wrote down the findings, including the one I can't fix.

fundamentals2026-06-29

Suggest, don't act: approval queues and the table that can't forget

Why high-stakes AI should propose rather than act, and what a trustworthy approval queue needs underneath it. The build is a Postgres audit table where UPDATE and DELETE fail by trigger, walked through line by line.

fundamentals2026-06-26

Give the agent a room, not a building

Agency helps when the path can't be scripted in advance, and it stays safe when the action space is designed so the worst possible run is affordable. The tiered mutation layer and the growth agent that have been teaching me this in production.

fundamentals2026-06-24

Context is a budget, not a backpack

Context engineering: million-token windows make stuffing everything in feel free, and it isn't. Attention degrades in the middle, and tokens cost the same whether they changed the answer or not. Three context strategies for the same task, with the real cost levers, batch and cache, priced out.

evals2026-06-21

The newest model lost, and only the eval knew

I benchmarked a range of vision models on the same real task and the newest, biggest one lost to an older, cheaper one. Without a small set of real outcomes to grade against, I would've shipped the wrong choice and never known. A field note on evals and golden sets.

ai-product2026-06-18

You don't need an ML engineer. You need a product engineer who doesn't trust the model.

Most companies shipping AI features aren't doing machine learning, they're doing product engineering around a probabilistic component. Hiring as if those are the same job is why so many AI features stall before production. On what the work actually is.

fundamentals2026-06-14

Why models fabricate, mechanically

Why language models fabricate, mechanically: the distribution has to put its mass somewhere, and 'I don't know' is rarely the most probable continuation. A temperature demo you can drag, what the research says about incentives, and the autopsy of my own four fabricated citations.

verification2026-06-14

The average answer is a trap

In a regulated domain the generic rule is usually close and occasionally, expensively, wrong. A system that answers from the average case is a confident-wrong machine. On why the unit of correctness is the specific instance, not the category.

fundamentals2026-06-10

Structured extraction: the validator is the product

Why output validation belongs outside the model, and what schema-constrained extraction looks like when the input is thousands of messy PDFs. The schema is a contract, the validator enforces it, and fields the model can't fill stay empty instead of guessed.

verification2026-06-07

The most useful thing my tool does is refuse to answer

A readiness checker that confidently flags the wrong thing is worse than no tool at all. The feature that makes mine trustworthy is the one that says 'I can't confirm this, check it yourself.' On designing abstention in from the start.

fundamentals2026-06-06

The queries vector search can't see

Identifier queries like 'O. Reg. 176/26' are where pure vector search goes blind, because rare exact strings have no semantic neighborhood. A live BM25 demo with per-term scores, why my pipeline keeps a full-text index next to its vector index, and where a reranker would earn its cost.

product2026-06-03

Stop predicting, start verifying

PermitCheck started as a tool to predict whether a permit would be approved. That promise was unkeepable, and no amount of data could fix it, because the outcome didn't live in the document. The fix was to change the question. On the difference between a prediction you can't keep and a verification you can.

fundamentals2026-06-02

Chunking decides what your AI can know

Chunk boundaries dominate retrieval quality more than model choice. A live visualizer shows a height limit getting separated from the exception that overrides it, and what that failure looks like in legal text.

verification2026-05-31

Ontario changes its laws about 500 times a year. I built the tool that reads every one.

Enacted tracks all 3,055 Ontario regulations and statutes, computes the exact diff when one changes, and explains it in plain English. The interesting part is what the AI isn't allowed to do.

fundamentals2026-05-29

What cosine similarity actually measures

Cosine similarity is an angle between two points in a space built to capture relatedness, and relatedness is not assertion. A draggable demo of the geometry, the real pgvector setup behind my research pipeline, and the case where I chose no embeddings at all.

verification2026-05-28

Never let the LLM author the numbers

A hallucinated macro total is a lie someone eats. Every number in my systems comes from plain code, and the model only chooses between validated options.

fundamentals2026-05-25

Where the arithmetic actually lives

Week 2 of going deeper on production AI fundamentals: which numbers the model should never author. The primary sources say models predict tokens, not sums, and the field's own answer is to hand the math to a tool. A walk through the four-layer resolver behind my meal engine, and why the LLM sits at the bottom of it, flagged.

pipelines2026-05-24

Extracting structured claims from 11,000 papers on a desk, not a datacenter

The architecture of a research extraction pipeline that has triaged 90,111 papers across its two big corpora and fully extracted 11,490 of them: acquisition, schema-constrained extraction, entity resolution, and the optimizations that mattered more than model choice.

verification2026-05-21

My research agent fabricated 4 of 5 citations, so I built a verification gate

I spot-checked the citations my extraction pipeline produced and most of them were wrong. Here's the gate I built so it can't happen again, and why I now wrap deterministic checks around every probabilistic system I ship.

FUNDAMENTALS

The fundamentals, rebuilt

12 weeks · primary sources

How do you wrap deterministic checks around a probabilistic system so fabrications cannot ship?

The build: A citation verifier that validates every cited source against an authority API

Read the essay

Which numbers in an AI system should the model never author, and where does the arithmetic actually live?

The build: A deterministic calculation layer the LLM can invoke but never override

Read the essay

What does cosine similarity over embedding space actually measure, and which queries does it quietly fail?

The build: Embed one corpus two ways and show where the retrievals disagree

Read the essay

Why does chunk size and overlap dominate retrieval quality more than model choice?

The build: Same corpus, three chunking strategies, measured hit rates

Read the essay
All 12 weeks →

FAQ

Asked and answered

—— The questions people actually ask about the work, all in one place.

I build complex product interfaces and the systems behind them. My strongest tools are React and TypeScript. My current public work focuses on AI and data pipelines with verification gates, eval harnesses, and audit trails.

Because my own pipeline once fabricated 4 of 5 citations and nothing about the output looked wrong. If a property of the output must always hold, something other than the model has to enforce it. That principle shapes everything I ship.

A 12-week public curriculum: each week I take one fundamental behind systems I already run in production, read the primary sources, rebuild it small enough to break on purpose, and publish the explainer I wish existed.

TypeScript and Python, Next.js, Postgres, and whichever model fits the task. The interesting decisions are rarely the model: they are chunking, retrieval, schemas, evals, and where the deterministic checks live.

Yes. Everything ships to the writing section and the email list as builds finish. The projects pages show the verification story behind each system.

How I work

Verified by default, explained in public.
Every system I ship has a verification story: what must always hold, and the deterministic check that enforces it.

28,000+ papers extracted, 333,000 triaged

Every citation machine-verified against OpenAlex + Crossref.

Read the essays
Spotlight
“Four of the five were wrong. … fabricated citations look exactly like real ones. … So I stopped asking the model to be trustworthy and built a gate.”

The verification gate

Week 01 of Production AI Fundamentals

Read it

One useful essay a week

What breaks in production,
written down while it's fresh.

Build logs and failure reports from systems where the output has to be right. No roundups, no reposts. Unsubscribe takes one click.

Joins the same list as the footer form · nothing else is sent