Independent · No vendor briefings · Paid at retail Methodology · Dispatch · RSS
About NEURALPOST

Run by two researchers, not a media holding company.

We started NEURALPOST in 2024 because every AI tool review we trusted came from a Hacker News thread, and every review that came up first on Google seemed written by someone who'd never used the tool past the free trial. This is our attempt to fix that, at a small scale.

Who we are.

Chen Wei is the lead researcher. He's shipped three production AI systems, has benchmarked over 40 LLM APIs, and runs the model, server, and developer tooling reviews. He's based in San Francisco and has strong opinions about cold-start latency.

Sara Kim writes about AI writing tools, creative applications, and the human side of working with language models. Former technical editor, current prompt engineer. Tests anything you'd use to generate or refine text. Based in Berlin.

What we tested in 2026 so far.

47 Tools benchmarked
2.3MTokens processed
840Hours of testing
14 Guides published
CW

Chen Wei

Lead researcher · San Francisco, CA

LLMs, servers, developer tools. Has benchmarked every major API at least twice.

SK

Sara Kim

Research editor · Berlin, DE

Writing tools, creative AI, agent workflows. Won't review anything she hasn't used on a real deadline.

Our principles.

The short version
01

We pay for everything at retail.

No vendor briefings, no free trials extended in exchange for coverage, no ambassador credits. Every tool in our reviews was paid for through the same accounts you'd use, which keeps the incentives clean.

02

Months of use, not a weekend sprint.

A tool has to be in our active workflow for at least three real projects before it's eligible to be written up. Most stay in rotation for a full quarter.

03

We say what didn't work.

Every guide names the tools that lost. Pros and cons aren't a both-sides ritual — they're notes from real failures on real work.

04

Two researchers, named.

No anonymous reviewers, no rotating contractor pool. Every guide is signed; both researchers have been building with AI tools for years.

05

We don't take placement money.

Vendors cannot pay to appear on this site. Not in a sponsored slot, not as a 'partner spotlight', not bundled with anything.

06

We update guides when things change.

When a tool degrades — pricing hike, model downgrade, API reliability drop — we say so and resort the list. The update date isn't decorative.

How we make money.

Affiliate commissions on the tools we recommend. The vendor pays us — never you. We disclose this everywhere, including on every link.

What we never do.

Accept payment for placement. Run sponsored content disguised as reviews. Take free extended trials in exchange for coverage. Recommend tools we wouldn't pay for ourselves.

What you can do.

If you want to support us: subscribe through our links when you'd subscribe anyway, or sign up for the signal. We've never asked for a tip.