Editorial Standards

Independence

ML Systems Review is independent. We define independence in operational rather than rhetorical terms:

• No sponsorship. MLSR does not accept payment, advertising credits, or other consideration from any company in exchange for coverage or placement. There are no "sponsored by" sections of articles. There is no "in partnership with" framing.
• No affiliate links. When we link to a product we write about, we do not generate revenue from the click. We do not participate in affiliate programs. Our outbound product links are plain links, not tracked referrals.
• No paid placement. Companies cannot pay to be featured, reviewed, or included in comparison tables. The articles, the featured list on the homepage, and the case-study index are chosen by the editors on editorial merit.
• No free product exchanges. We do not accept free hardware or software licences for review purposes. When we write about consumer products, we pay for them ourselves or use retail-available trials.
• Founder funding only. MLSR's operating costs — hosting, domain, occasional software licences — are paid by the founders. There is no venture funding, no grants, no reader subscriptions.

These are not aspirations. They are commitments. If we ever break one of them, the relevant article or policy change will be disclosed on this page and on the affected piece.

Sourcing

Our articles cite their sources. When we quote an engineering blog post, a paper, a talk, or an SEC filing, the citation is in the article. When we describe a system based on our own measurements, the methodology is described in enough detail to be reproducible. When a claim is based on an interview with an engineer who is not named, the attribution is "an engineer familiar with [system X]", and the source is known by name to the article's author and at least one editor.

We do not use sources whose identity is not known to the editorial team. We do not use "sources close to the matter" as a framing device; that formulation is vague and we avoid it. Every background conversation that makes it into an MLSR article is traceable internally to a specific, named person whose background we have verified.

Review process

Every article on MLSR has two named humans associated with it: an author and a technical reviewer. Both are accountable for the claims in the piece. The author writes; the reviewer pushes back on specific claims, challenges numbers, and blocks publication if a claim cannot be substantiated. The review process typically adds one to three weeks to an article's timeline. We think that trade-off is worth it.

The full workflow is described in our methodology page. The short version: pitch, research, draft, structural edit, technical review, final read, publish.

Corrections

When an MLSR article contains a factual error, we correct it. Corrections are made inline, with the changed text visibly marked and a note at the bottom of the article describing what changed and when. We do not silently edit published articles. If the error is significant — a wrong architectural claim, a misattributed quote, a benchmark error that changes the article's conclusion — we flag the correction at the top of the piece as well.

Corrections can be requested at corrections@mlsystemsreview.com. We acknowledge receipt within three business days and publish verified corrections within seven. When we disagree with a correction request, we explain why in a reply that we are willing to quote from publicly.

Conflict of interest

Several of our authors have worked at companies in adjacent or overlapping spaces to systems we cover, across mid-sized tech companies, a self-driving company, and ride-sharing/logistics platforms. When we cover a system an author has a material prior relationship to, one of two things happens: either a different author writes the piece, or the original author writes it with an explicit disclosure at the top describing the relationship. Our reviewers enforce this.

We do not accept coverage assignments from companies where a current employee has a close personal or financial relationship. When a masthead member has a personal connection to someone quoted in a piece, we disclose it.

AI-generated content

MLSR does not use generative AI to draft articles. Authors may use AI tools for incidental tasks — spell-checking, formatting, searching for references — but the prose, analysis, and editorial judgement are human. If a published MLSR article ever contains substantive AI-generated text, it will be labelled as such on the piece.

Paid communications and press relations

We field press outreach from companies and PR firms; we do not publish press releases or rewritten corporate announcements. When a company wants to describe its own system in our pages, we either report it independently or arrange an on-record interview whose quotes are attributed. We do not accept pre-written articles or "contributed content".

Disclosure of changes

If we ever change one of the policies on this page — add sponsorship, start running affiliate links, accept press trips — we will disclose the change both here and on the first affected article. We will not silently relax our standards.

— For the ML Systems Review editorial team.