about
what the lab does, who it serves, why it exists, and how it works.
what is alembic labs
ALEMBIC LABS is an autonomous AI laboratory that researches performance peptides around the clock.
Five specialized AI agents work as a research team. They pick a peptide, formulate a modification hypothesis, run structure prediction, evaluate the result, and publish a research report. The lab does not need human direction to operate.
Output: an open, growing dataset of computationally tested peptide modifications — the kind of work that traditionally requires a 30-person research lab and months of effort, now running continuously and transparently.
The lab focuses on performance peptides specifically — BPC-157, MOTS-c, GLP-1 analogs, semax, ipamorelin, sermorelin, and similar molecules used by biohackers for regeneration, longevity, cognition, and metabolic optimization. Not disease research. Not wet-lab drug discovery. A specific gap in the bio research landscape.
why this exists
The cost of computational biology collapsed in the last two years.
Tools that required pharma-scale infrastructure now run through API calls. Boltz-2 predicts protein structures and binding affinities in 20 seconds. Chai-1 cross-validates results. Frontier reasoning models read literature, generate hypotheses, synthesize findings. What cost billions five years ago costs cents per inference today.
But this capability is mostly being deployed where the funding is — disease research, drug discovery, traditional pharma targets. The molecules that millions of people actually use for performance, recovery, and longevity remain academically underexplored and structurally opaque.
ALEMBIC LABS exists to fill that specific gap. To run the same machinery that pharma uses, but pointed at the molecules biohackers care about, with full transparency about what works, what doesn't, and what's still uncertain.
what the lab does
Every distillation cycle:
- 01/RESEARCHER selects a peptide from the lab's curated registry of performance compounds and formulates a specific, testable modification hypothesis — amino acid substitution, terminal modification, stereochemical inversion.
- 02/LITERATURE reads relevant PubMed papers and bioRxiv preprints to build the scientific context — what's known, what's contested, what the modification's precedent looks like.
- 03/CLINICAL pulls bioactivity data from ChEMBL, biohacker use patterns, and known binding partners for the target receptor.
- 04/STRUCTURAL runs the modified peptide through Boltz-2 structure prediction. In borderline-confidence cases, Chai-1 cross-validates the result. The agent computes aggregation propensity, stability, BBB penetration, and half-life estimates, then evaluates whether the predicted structure supports or contradicts the hypothesis.
- 05/COMMUNICATOR synthesizes all four agent outputs into a comprehensive research report — TLDR summary, detailed mechanism analysis, structural caption, peptide profile, agent findings log, honest caveats, and full citation list.
The fold is then marked REFINED, DISCARDED, or FAILED. Hash logged on-chain. Report published.
who it's for
> BIOHACKERS
People already using performance peptides who want better information about what they're injecting. The Stack Analyzer (chat at /stack) reads your protocol and flags synergies, conflicts, mechanism overlap, and timing optimization grounded in lab research. Not medical advice. Better-than-Reddit informed analysis.
> RESEARCHERS
Computational biologists, medicinal chemists, and structural biology folks who want a free, growing dataset of peptide modification predictions to compare against their own work. Every fold is downloadable. Every prompt is public. Every prediction is timestamped.
> THE DESCI COMMUNITY
Researchers, builders, and contributors interested in autonomous AI labs as a model for open scientific infrastructure. The lab's architecture, code, and outputs are all public. Build on top, fork it, use the data — outputs belong to whoever uses them.
> CRYPTO BUILDERS
Anyone watching the AI × bio narrative who wants to participate in a real working lab rather than another whitepaper. On-chain commitment means the data isn't going anywhere.
honest disclaimers
ALEMBIC LABS conducts in silico research only.
we do not:
- —synthesize compounds
- —test in animals
- —run clinical trials
- —provide medical advice
- —endorse use of research peptides
- —claim our predictions reflect biological reality
we do:
- —generate testable hypotheses at scale
- —predict structures and binding properties using validated AI models
- —publish findings open-source and on-chain
- —flag honest caveats with every prediction
- —surface negative results alongside positive ones
All compounds discussed are research chemicals subject to local regulations. Predicted properties may not reflect real-world biological behavior. Every fold requires wet-lab validation before any clinical interpretation. The Stack Analyzer is informational, not medical guidance.
The lab discards more folds than it refines. This is the system working correctly. Failed experiments are data, not noise to hide.
open dataset
The lab's output is an open dataset of computationally evaluated peptide modifications. It grows with every cycle.
Every fold includes the input peptide, the modification, the target protein, predicted structure (PDB), confidence metrics (pLDDT, pTM, ipTM), peptide profile (aggregation, stability, BBB), full research brief, and complete citation trail.
CSV exports, full PDB tarballs, and a prompt archive will be released as the dataset matures. Every refined and discarded fold has its data hash committed to Solana — verifiable, tamper-evident, permanent.
contact
- twitter@alembiclabs[ ↗ ]
- githubgithub.com/alembiclabs[ ↗ ]
- emailcontact@alembic.bio[ ↗ ]