The transparency page is honest. The homepage headline is not — and there's no named operator to hold to either one.— CrazyCheck.ai editorial, June 2026
What BioLLM actually is
Visit biollm.com and the first thing you read is "The World's First Living Language Model — BioLLM bridges biological neural intelligence with advanced language modeling, creating an AI that thinks more like you." It's a striking claim. Biological neural intelligence. A model that thinks. The implied promise is an AI with something qualitatively different going on under the hood — a step beyond the transformer architectures that power every major language model today.
Here is what is actually running: Qwen3.5-4B, an open-source model released by Alibaba Cloud under the Apache 2.0 license, fine-tuned on 24,000 stimulation trials captured from Cortical Labs CL1 multi-electrode array (MEA) experiments. In production, when you send a message, it is processed not by living neurons but by a 10-million-node Izhikevich spiking neuron simulation — a mathematical model developed by computational neuroscientist Eugene Izhikevich in 2003 that replicates the firing dynamics of biological neurons as differential equations. It is real technology. It is not biology.
To their credit, BioLLM's transparency page states this clearly: "BioLLM does NOT run on live biological neurons in production. Inference uses a 10 million Izhikevich neuron simulation." That sentence is accurate and important. The problem is that the homepage headline — the phrase most users will read and most journalists will quote — says the opposite.
- Branding: "World's First Living Language Model"
- Reality: Qwen3.5-4B + a 10M-node Izhikevich simulation — not live neurons
The operator: a Florida LLC with no face
The legal entity behind BioLLM is Stormy IT Services LLC, a Florida limited liability company doing business as Ankythera AI. This is disclosed in the Terms of Service. That is all that is disclosed. There are no named founders, no executive team, no LinkedIn profiles, no investor deck, no physical address, no email address on the terms page, and no public regulatory filings that would identify the humans operating this platform.
The sole public contact is @BioLLMbot — an automated bot on X (Twitter) and Telegram. WHOIS privacy is enabled on biollm.com (registered 2024-10-28 via Namecheap), so even the domain registration record reveals nothing. The domain is 20 months old, predating the March 2026 token peak, which rules out a pure opportunistic rush-launch. But age does not substitute for identity.
This anonymity matters for several reasons. The platform offers a genomic data tool (OpenCure) that accepts raw DNA files. It operates a cryptocurrency token ($BioLLM) sold to retail holders. It stores user data for "model improvement." Under each of these activities, the question "who is responsible?" has no answer that can be verified by any user of the platform.
- Entity: Stormy IT Services LLC (DBA Ankythera AI), Florida
- No named founders, no address, no email — contact is @BioLLMbot only
The Pump.fun token
The $BioLLM token is a Solana SPL token launched on Pump.fun (contract address: 9oxGicd5KXNcthYM6yD23X9zcznQVnHjrAEN38fFpump — the "pump" suffix is not a coincidence, it is the standard Pump.fun contract identifier pattern). Supply is 1 billion tokens. Mint authority is revoked. Tax is 0%.
Pump.fun is a permissionless Solana launchpad: anyone can create a token in seconds with no vetting, no compliance review, and no project requirements. Per Solidus Labs research published in 2024, 98.6% of tokens created on Pump.fun show at least one fraud signal — wash trading, coordinated developer wallet dumps, or structured rug pulls. Launching on Pump.fun does not prove fraud. It is, however, a launch context that provides essentially zero signal about project legitimacy and very strong prior odds of value decay.
$BIOLLM reached its all-time high of $0.002426 on March 8, 2026. At the time of this audit it trades at approximately $0.000811 — a 66% decline. Market cap sits around $811,000 with minimal volume. The mint revocation and zero-tax structure are positive parameters that reduce specific rug-pull vectors, but they do not change the trajectory or the launchpad context.
Users can pay for premium BioLLM subscriptions in $BioLLM tokens. This creates a functional loop in which the token's utility is circular: the platform sells tokens, tokens are used to pay the platform. There is no refund policy published for any subscription tier, including those denominated in a volatile asset.
- $BioLLM launched on Pump.fun (98.6% of its tokens show fraud signals)
- Down ~66% from its March 2026 ATH; circular token utility, no refunds
OpenCure: your DNA, their model
OpenCure is BioLLM's genomic health tool. The pitch: upload your raw genomic data file (.vcf format, 23andMe export, or similar), and receive "personalized health insights powered by BioLLM." The privacy policy states data is "not sold to third parties" but is retained for "model improvement."
Genomic data is among the most sensitive categories of personal information that exists. It is permanent, it is uniquely identifying, it reveals predispositions to disease, it can affect family members who never consented to sharing, and it cannot be changed if mishandled. Uploading it to any platform carries inherent risk. The risk calculus for OpenCure is extreme: the receiving party is a fully anonymous LLC with no named clinical professionals, no IRB (Institutional Review Board) approval process, no CLIA (Clinical Laboratory Improvement Amendments) certification, and no FDA regulatory framework cited for health-adjacent interpretation of genetic data.
The product may generate interesting AI-assisted summaries. The cost is trusting your irreversible biological identity to an operator whose human identity is unknown.
- Solicits raw genomic uploads (.vcf, 23andMe) retained for "model improvement"
- No named clinician, no IRB, no CLIA, no FDA framework cited
What actually works
The AI chat product is functional. The free tier (5 messages per day per IP address) requires no account, no wallet, no email. Qwen3.5-4B is a capable base model — its outputs as a chat assistant are on par with mid-tier consumer AI tools. The Izhikevich neuron simulation, while not "living," is a genuine computational neuroscience technique. Cortical Labs and the CL1 MEA platform are real — Cortical Labs is a Melbourne-based biotech with commercially available biological computing hardware. The fine-tuning methodology described (capturing spike-timing correlations from CL1 sessions as auxiliary training context) is scientifically coherent, not fabricated.
The Developer API and Staking sections of the platform are live (at varying levels of completeness). The transparency page is a genuinely honest document. These are not small credits — many AI-crypto projects do not disclose even this much about their technical stack.
- Free chat tier (5 msgs/day/IP) works with no wallet or email
- Transparency page is a genuinely honest technical disclosure
BioLLM Spark: the hardware that doesn't exist
The homepage lists "BioLLM Spark" as an upcoming product: "Record your neural patterns and have BioLLM learn from your unique brain." It is presented as an EEG headset that would connect a user's actual brain activity to the BioLLM model. There is no shipping date, no pre-order link, no prototype images, no FDA 510(k) pre-submission notice, no manufacturing partner named, and no independent coverage of the device from any publication at audit date.
BioLLM Spark appears, at present, to be a concept — a roadmap item dressed as an imminent product. Listing it alongside live features like Chat and OpenCure, without any "in development" qualifier, is a meaningful transparency failure.
- EEG headset listed as a live product with no ship date or prototype
- No FDA clearance, no manufacturer, no independent coverage
The bottom line
BioLLM is an interesting project wrapped in serious structural problems. The core AI product (Qwen3.5-4B fine-tuned on MEA data) is real and functional. The neuroscience framing, while overstated in the headline, has legitimate technical basis on the transparency page. The Cortical Labs data source is plausible.
But: the operator is completely anonymous. The token was launched on a launchpad where 98.6% of projects show fraud signals, and is now down 66% from its ATH. OpenCure collects irreversible genomic data with no clinical oversight and no named humans responsible. The EEG headset does not appear to exist. Payments are crypto-only with no refund policy.
The combination of a misleading headline, anonymous ownership, memecoin tokenomics, and genomic data collection is disqualifying for a Cautious verdict. This one is Avoid.
- Real, functional AI — but disqualifying structural risk
- Anonymous team + memecoin + genomic data collection = Avoid
Notable community threads
ATH $0.002426 (2026-03-08). Current ~$0.000811. Down ~66% from ATH. Market cap ~$811K. Launched on Pump.fun. Contract ends in "pump" suffix confirming launchpad origin. — $BIOLLM / SOL — DexScreener · dexscreener · negative