daimones
AI AssistantsUncensored Philosophical AI Chatbot
AI that thinks like a philosopher, not a corporate lawyer. Uncensored, source-grounded reasoning.
About
An AI that reasons dialectically instead of dodging. daïmōnes runs open-weight models (Qwen3.8-27B) on its own sovereign infrastructure, with retrieval over primary philosophical texts, so answers cite sources you can verify instead of hedged corporate boilerplate. Aristotle is the first persona. On the public Golden Benchmark v2.1 (Sept 2026, identical system prompt for every model), daïmōnes scored 84/100 vs ChatGPT GPT-5.4 at 79, Gemini 3.6-flash at 70.4 and Grok 4.3 at 64.7. A single $600 VM beat the frontier labs on philosophical reasoning. Built for students, researchers and academics who need philosophically literate answers, institutions exploring sovereign AI deployment, and anyone tired of assistants that lecture instead of think. No alignment theater. No engagement-optimized flattery. Free tier available. daimones.ai
From the founder
Vasileios StergiouWho is daimones for?
Students, researchers and academics who need philosophically literate answers grounded in primary sources, not summaries of summaries. Institutions exploring sovereign AI deployment, where data never leaves infrastructure they control. Semester pilots and department programs are scoped per institution. And anyone tired of assistants that lecture instead of think. If you have ever asked an AI a hard question about ethics, politics or meaning and gotten a hedged non-answer, daïmōnes was built for you.
What problem does daimones solve?
Corporate AI assistants are optimized to avoid offense, not to find truth. RLHF and alignment layers turn difficult questions into hedged boilerplate: the model apologizes, deflects, or lectures instead of reasoning. For philosophy this is fatal. You cannot study Aristotle through a filter that fears controversy. daïmōnes solves this by running open-weight models on its own sovereign infrastructure, with retrieval over primary philosophical texts. No corporate filtering layer sits between you and the model. Answers engage taboo and difficult questions honestly and cite sources you can verify. On the public Golden Benchmark v2.1, daïmōnes scored 84/100 vs ChatGPT GPT-5.4 at 79 on philosophical reasoning.
How is daimones different from the alternatives?
Three things. One: sovereignty. daïmōnes is not reselling an API. It runs open-weight models (Qwen3.8-27B) on self-hosted infrastructure, so no corporate policy layer can silently rewrite answers. Two: grounding. Retrieval over primary philosophical texts means answers carry citations, not vibes. Aristotle is the first persona, tuned to reason dialectically in the tradition of the Lyceum. Three: proof. Golden Benchmark v2.1 (Sept 2026, identical system prompt for every model): daïmōnes 84/100, ChatGPT GPT-5.4 79, Gemini 3.6-flash 70.4, Grok 4.3 64.7. Methodology and scored results are published. A single $600 VM beat the frontier labs.
How much does daimones cost?
Freemium, no credit card required. Observer (free): 3 messages per day with Aristotle. Disciple ($29.99/month): unlimited dialogue, full conversation history, academic research tools (tagging, export, search), Daily Philosophical Challenge, mobile access, email support. 14-day free trial. Archon ($99.99/month): everything in Disciple, plus Citation and Grounding Reports, Lecture Mode, early access to new philosopher personas, priority response queue, founder recognition. Universities, NGOs and think tanks: bespoke deployment programs, quoted per institution within 48 hours via [email protected]
What's next for daimones?
More personas. Aristotle is the proof of concept; Plato and Marcus Aurelius are next on the roadmap, each grounded in its own primary-source corpus. Voice mode. Work has begun on streaming text-to-speech so Aristotle can answer aloud, using open-weight TTS running on our own hardware. Institutional deployment. Semester pilots and department programs let universities run daïmōnes for coursework, with data staying inside infrastructure they control. Continued benchmark transparency. Every Golden Benchmark run is published with a full methodology, so our claims stay verifiable. Follow the blog on our website.
Startup Battle record
daimones went 2–11 (15% wins) in the Startup Battle in week 34, where visitors pick the better of two startups head to head. It faced 9 different startups.
- Split withTeenCycle1–1
- Lost toAllowanceGo2–0
- Lost toBali Peptide Lab2–0
- Lost toHuntoso Privileged Access Management2–0
- BeatTabTotaler1–0
- Lost toAgentplace1–0
- Lost toCommercial Use Font1–0
- Lost toracinglines1–0
- Lost toTranscribeYT1–0
October 27, 2026Week 35
Vasileios StergiouReviews
Comments
Sign in to leave a comment
Sign In