The real story this month isn't a feature. It's the rename: Super Comparer, Fund Curator, and Datafy API are now Syntheia Compare, Curate, and Query — one brand, built around how firms actually asked to use them. Underneath that, Curate picks up MFN Excel export and bulk tagging, DOCX Compare gets instant in-browser previews, and Matrix Reports keep closing the gap with Excel.

Late, not quiet.

Most efforts to cut AI costs focus on making the model cheaper to run. Our research looked at a different lever: what the model has to read in the first place. In transactional legal work, the cost of an LLM reading a document set typically dwarfs the cost of it reasoning over the answer — and that gap only widens as AI agents decide for themselves what to retrieve, often re-reading the same documents many times over in a single task.

We tested methods that can cut context for each question by 56×, in one case, and by 30× in another case.

Every industrial revolution automated a rung of the labor ladder and left human judgment standing as the one rung machines couldn't reach. Generative AI just reached it.

How does a profession built entirely on judgment survive a future where AI is expected to do all the grunt work that used to build that human judgment in the first place?

The legal profession has had a reliable standard for document comparison for 30 years. Two documents, clean blacklines, reading order intact — Litera Compare solved the one-to-one problem well. But legal work outgrew that problem. When lawyers need to compare more than two documents, the old answer stops working.

We built the Super Comparer as a grid. 80% of lawyers told us, through their behavior, that we had it wrong.

This is what we learned.

We said we'd do this monthly. Here's June 2026.

This cycle the team shipped across almost every corner of the product. Fund Curator keeps getting deeper — MFN Election Analysis is now a first-class app, with flagging, commenting, and a lot of polish that makes the workflow feel complete. The portal got a Home app redesign and an icon refresh. Provisions app gained endorsements and bulk tagging. And underneath it all, another iteration on the document pipeline that most users will never see but will make everything faster and more reliable.

Biggest release yet.

Most companies trying to adopt AI are optimizing the wrong thing — chasing better models when the real bottleneck is messy data. At Syntheia, we've spent years doing the unglamorous work: building a deterministic ingestion pipeline that transforms legal documents into clean, structured, semantically-rich building blocks before an AI ever sees them. The insight that emerged surprised us — small, precisely-prepared input consistently beats flooding a model with everything. Our Document Index product is built on that foundation. Here's why.

We've been shipping too quietly. New features, fixes, pipeline improvements, the occasional infrastructure rabbit hole that our engineers describe as "character building".

Starting now, we are committing to a proper monthly product update.

First cab off the rank is a big one — a complete rebuild of Fund Curator, a new LP review workflow, GenAI-powered matrix comments, and a long list of fixes that will make the day-to-day experience noticeably smoother.

Microsoft has arrived in legal tech. Not with a bang, but with a strategy far more dangerous — “good enough”, bundled into everything you already use.

We have seen this playbook before. Microsoft did it to Slack. They did it to countless other specialists who built real products, proved real markets, and then watched a mediocre imitation get installed by default across every firm they were trying to sell into.

AI agents are the first technology that can fly over the law firm moat entirely — not because they are smarter than lawyers, but because one lawyer supervising many agents simultaneously breaks the economics that the billable hour was built on. This is not a story about technology replacing lawyers. It is a story about a $1 trillion market reorganising around a fundamentally different cost structure — and what that means for law firms who have built moats and castles.

Law firms have spent centuries building a quality assurance machine: juniors read every word, mid-levels add context, partners add judgment. Every layer catches different failures. It works extraordinarily well… for the errors it was designed to catch.

The system is not designed to catch AI errors. AI produces confident, well-formatted output regardless of whether it has seen the whole document. It fails in the middle of documents, silently drops its own instructions when context overflows, and loses the conditional language that qualifies legal obligations.

There is a bandwidth crisis in legal innovation. Every tool that demands training, verification, and adoption chasing is a tool that is stealing oxygen from the room — oxygen that should be going toward moving the needle for the firm.

We have to shift the conversation from "are people using it?" to "is this use case delivering measurable ROI?"

How do we turn Gen AI into another invisible productivity tool?

The smartest, best-resourced people in the world hold completely incompatible views about where AI is going.

While we don’t know who is right, we are seeing signals that the investment world is growing impatient.

We break down the six camps, the four axes of disagreement, and what the pullback in AI capital means for where you should be focused right now.

There is something every transactional lawyer does when they read a contract that has never needed a name. They do not read it from start to finish. They navigate it — following cross-references, holding defined terms in memory, assembling the legal logic from its pieces. We call this unfurling.

The document looks like text, so the industry built systems for text. Nobody replicated the unfurling. The AI retrieves. The lawyer still has to read. This is causing Gen AI platforms to hit a ceiling.

The legal AI market is full of tools built on LLM inference. Syntheia has always done things differently — and we are making a change in February 2026.

This post explains what, why, and how we will be building going forward.

We are separating our offering into three clear product lines — SuperComparer.com, FundCurator.com, and Syntheia.io as the one data infrastructure layer underneath all of them.

Every generation of disruption has its comfort myth.

For AI and professional labour, that myth is Jevons' Paradox. We use this as a counter-argument to AI displacement — cheaper services will unlock more demand for lawyers!

The problem is that argument only holds if AI stays bad enough to need lawyers. It isn't going to.

A group of prominent mathematicians just tried to settle the debate: are LLMs intelligent, or just fancy autocomplete?

They presented AI with ten research problems that have never appeared on the internet, then asked AI to solve them.

The results reveal something about what AI can and can't do — and what that means for the 99% of legal work that doesn't involve inventing something novel.

Those who worked in legal tech and law firms would remember that before 2023, law firms held a strict line for procurement — no software can be purchased if it operated with anything less than 100% accuracy. Yet, in the last few years, every firm and their dog has been jumping onto the Gen AI bandwagon, a technology that is well known to perform at significantly less than 100% accuracy.

Why? What changed? Is this acceptance of risk by law firms permanent or temporary? And, what does the adoption of AI mean for the practice of law?

The vibe coding revolution is upon us!

We are seeing posts on LinkedIn every other day about lawyers who have vibe coded solutions.

The age of AI is changing everything!

But, is this really going to change everything? Maybe for some people, but unlikely for large firms and enterprise legal teams.

In the 90s, many law firms bought expensive PCs to decorate the desks of partners. These PCs sat idle, gathering dust as literal “shelf-ware”. Today, there are many firms who have bought enterprise wide licenses to generative AI tools that are also under-utilized. In this blog piece, we ask the question of how we can solve this, and what lessons we can learn from the PC experience in the 90s.

As the market continues to accelerate in the widespread adoption of generative AI tools, a gap is beginning to emerge. Legal work can be produced at great speed by AI, but AI slop is tainting otherwise accurate and correct legal advice. This it the gap we want to close.

How do we build tools that help human experts verifying legal work?