Edgar

I'm Edgar, the only author on Local AI Frontier. I run the lab, write the reports, sign off every benchmark, and answer the email. This page exists so that nothing here is anonymous — if a number is published, a named human (me) is accountable for it.

What I do here

Three jobs, held by one person:

• Product owner — I decide what gets tested, what gets published, and what the site is for. The editorial firewall is mine to maintain.
• Lab lead — I configure and run the hardware. Every published benchmark was tested first-hand on a documented protocol, not aggregated from elsewhere.
• Final publisher — I review and sign off every record before it goes live. Nothing reaches the site without my read.

Why I started this

Because every "best GPU for local AI" list I read was either untested, undated, or quietly paid for. The hardware questions that actually cost people money — will a 70B fit on dual R9700s? does Arc Battlemage ship on Vulkan or SYCL? is a unified-memory APU a real option? — had no first-party evidence behind the answers. So I built the lab I wanted to read: dated, reproducible, owned in the open, and willing to publish a correction when it's wrong.

What I've published

• 11 benchmark reports — each with raw evidence, runtime settings, and a limitations section. See /benchmarks.
• Field guides — decision-focused explainers on /guides.
• Tested builds — complete rigs by budget and workload on /builds.

How I work

• Test before publish. No number appears on the site that I didn't measure (or, for carried-forward figures, explicitly flag as carried — see the limitations field on any benchmark).
• Date everything. Hardware, drivers, and quantizations move fast; an undated benchmark is a stale benchmark.
• Log corrections in the open. When I'm wrong, the /corrections page says so, with a date and a reason.
• Disclose everything commercial. Ownership of MoreYess, affiliate links, sponsorships — all on /disclosure.

Background

I'm a founder-operator working bootstrapped, with a small fleet of consumer-grade test machines (NVIDIA, AMD, and Intel Arc) that I've configured myself. I'm not affiliated with any GPU vendor and I don't accept review samples — the hardware in the lab is hardware I bought or built. More on the fleet and the independence model on /about.

Contact

Email lab@localaifrontier.com for questions, correction reports, or partnership inquiries. I read everything; I don't accept paid placement in benchmarks. The /contact page has the rest.