@fijnmin
@fijnmin
Founder/Developer at TruthSignal AI
Who Is This Person?
Philip Marais, operating under the Twitter handle @fijnmin, is actively involved in developing and promoting TruthSignal AI, a Bayesian-powered tool for credibility analysis of factual claims and predictions. The tool leverages large language models such as Grok, Claude, and Perplexity to assess truthfulness by incorporating base rates as priors and conducting comprehensive research to build posteriors. Recent activities as of November 2025 include analyzing high-profile claims, such as Elon Musk's prediction on UK civil unrest (deemed 90% likely tribe signaling), verifying user posts, and demonstrating the tool's capabilities through Twitter interactions. Marais engages with communities around Polymarket, AI ethics, and fact-checking, tagging @truthsignal_ai for analyses and teasing premium features.
How Credible Are They?
Philip Marais (@fijnmin) appears credible as an emerging figure in AI-driven fact-checking, with activities centered on building and demonstrating a novel Bayesian truth verification tool. The absence of controversies, coupled with methodical approaches in public posts, supports high reliability in shared information. However, as a promoter of his own tool, outputs should be cross-verified for objectivity. Overall, strong potential for expertise in probabilistic analysis, though influence remains limited due to small-scale engagement.
Assessment by Grok AI
What's Their Track Record?
No prior fact-checks, corrections, or controversies identified in searches. Marais's posts demonstrate consistent promotion of transparent, evidence-based analysis via TruthSignal, with examples including probabilistic assessments of claims rather than absolute statements. Historical tweets align with truth-seeking themes, showing no patterns of misinformation or disputes.
What Have We Analyzed?
Recent posts and claims we've fact-checked from this author
Post by @fijnmin
@fijnmin · 22h ago
The other day it did something that really blew my hair back. Not routine by any means but pretty impressive. I was looking a house for sale, without a street address but neighbourhood listed and vague details of the location, including photos. I had Atlas do a prediction using a Bayesian approach and then put the location down in Google maps where the house was most likely to be. Research of listings returned no address. It was a statistical approach. Put the pin down, literally in front of the correct house. I was actually blown away.
Post by @fijnmin
@fijnmin · 4d ago
I have considered doing this for medical aid payments and reimbursements in South Africa, but I am conflicted. It is a nuanced situation because in SA, we have a set of minimum prescribed benefit for which the medical aid must legally pay. Now I am not suggesting impropriety on the side of medical aid. I am suggesting partially misaligned incentives. From my experience working in the insurance space, a medical aid is incentivised to reduce expensed, primarily in order to reduce something called benefit erosion. Basically, medical expenses inflation exceed premium collection and the only way a medical aid can offer lower premium inflation is to reduce the benefits for which the pay or to value of the contribution they make. This protects their premium inflation in order to increase market share or increase medical aid uptake of young members. Young members claim less and that is how they fund expenses for older members. This means that medical aids are very deliberate in their process of deciding authorisation and reimbursements with the aim of paying only what they need to. I have heard of a number of individual cases where members have engaged the medical aid on claims which have been denied. If we make it simpler for the members of a medical aid to protest specific payments, members will likely use such a service, as it will reduce their out of pocket expenses in the short term, but may in the long term inflate their monthly premiums, as the funds have to come from somewhere. I guess, the question is, should be build something to ensure that members are getting the maximum benefit from their medical aid, it may in the medium and long term reduce the contributions from medical aids and increase the out of pocket expenses for members. Another caveat is that the private medical industry can legally ask any price for a private service, but in most cases it is benchmarked against the Discovery rates. I am keen to hear thoughts on this.