Independent verification is a conversation, not a verdict handed down from one building. These are the people whose work on AI evaluation, safety, and reliability sharpens ours: the writers and researchers we read to keep ourselves honest.
To be clear up front: this is editorial. The people below are independent voices we admire. None of them endorse, advise, or are affiliated with us. We're simply pointing to what they made.
Arvind Narayanan & Sayash Kapoor, AI Snake Oil (Princeton). Among the clearest voices separating real AI capability from hype; their discipline of demanding evidence is the whole reason independent evaluation matters.
Rumman Chowdhury, Humane Intelligence. Pioneering work that treats red-teaming and algorithmic accountability as something you actually test, not just assert.
Helen Toner, CSET, Georgetown. Rigorous, sober analysis of how independent oversight of frontier AI can actually function.
Miles Brundage, independent researcher writing on the policy of frontier AI. On the gap between what labs can measure internally and what the public can actually verify.
Markus Anderljung, GovAI. Research on the practical mechanics of AI governance and evaluation, and how the abstract principles become real processes.
Nathan Lambert, Interconnects. Clear-eyed technical writing on open models and how we actually measure them.
Dr. Seth Dobrin, Silicon Sands News. On why smaller, specialized models, and verifiable ground truth, beat scale-for-its-own-sake, and on taking the reliability and hallucination problem seriously.
Gary Marcus, Marcus on AI. The field's most persistent skeptic; a useful counterweight whenever the discourse drifts toward hype.
These are independent voices we read and admire. Listing them is editorial commentary. It does not imply they endorse, are affiliated with, advise, or have reviewed The FixAI Group or ReallySolved. All names and works belong to their authors; every link goes to their own publication. If you're named here and would prefer not to be, contact us and we'll remove you promptly.
If you work on independent AI evaluation, safety, or reliability, or you just want to compare notes, we'd genuinely like to hear from you.
Contact us →