Nothing behind
closed doors.
Every protocol, dataset, model, and result is published under a permissive license.
An independent, open research institute
Science,
in the open.
Big questions. Rigorous experiments.
Knowledge that belongs to everyone.
Automated discovery. Independent critique. All public.
CONNECTED BY CURIOSITY
Energy · Compute · Biotech · Robotics
01 Our conviction
The next era of science
should belong to everyone.
Science is getting faster. The ability to check it needs to keep up. We are building an automated research lab in energy, compute, biotech, and robotics—with the methods, data, models, and failures open to all.
What we shareNothing behind
Trust comes
from the evidence.
An independent critic challenges each result before release. The author never reviews its own work.
Even a failure
moves us forward.
Failed replications enter the record. Recurring mistakes become checks for the next experiment.
02 Research programs
Four frontiers.
One shared platform.
From energy and computing to human biology and machines that act in the world. Four connected programs, built on the same open research infrastructure.
The research frontier
Fusion and grid systems
- Replicate published plasma-control and reactor-simulation results
- Version surrogate models so labs can re-run them unchanged
- Physical loop: partner tokamak and materials facilities
All four programs begin in simulation. The first fully funded vertical leads; the others start as pilot replication programs.
03 Our approach
An open record.
A continuous loop.
Six stages turn a question into knowledge anyone can build on. Explore each step, from the first hypothesis to the next discovery.
The research loop01 / 06
Start with a question worth checking.
Scientists select published results for replication. Every hypothesis begins with a traceable source and a clear question.
OutputA testable hypothesis
Led by scientists. Humans set the questions, approve physical protocols, and adjudicate disagreements.
Starting in simulation. Compute closes the first loops. Partner labs bring physical validation from year two.
Nothing exists until the registry versions it
Every dataset, model, protocol step, and result carries a tagged, versioned identity. Protocols reference tags, never prose. An unversioned input is a build error.
Protocols are assembled, not paraphrased
Locked components are inserted verbatim by string substitution. No model rephrases a method “for context.” Paraphrase is where silent drift enters.
An independent critic reviews before release
A separate model, in a fresh context with no memory of the work, attempts to break every result. The author never reviews its own output.
Every failure becomes a check
Diagnoses that recur are promoted to automated pre-execution checks. The lab never pays to rediscover the same mistake, and the check library is itself published.
04 The public record
The work.
Wide open.
Essay / 01
EnergyAIMaterials
The Exponent Is the Price: Quantum Tunneling and the Architecture of Wireless Electric Grids
The phrase "wireless power transfer by quantum tunneling" is, read literally, a category error. Single-particle electron tunneling through a vacuum or dielectric barrier is suppressed as …
September 1, 2026 · 34 min
Essay / 02
RoboticsAIEconomics
Cybercab, Robotaxi Fleets, and the Economic Structure of Autonomous Mobility
The dominant public argument about autonomous ride-hailing treats the driver as a cost line to be deleted. This paper argues that the driver is not a cost line but a bundle of four distinct…
September 1, 2026 · 35 min
Essay / 03
BioTechChemistryMedicine
Target Detection as the Rate-Limiting and Specificity-Limiting Step in CRISPR Genome Editing
CRISPR-Cas genome editing is usually described as a cutting technology. It is more accurately described as a search technology with a cutting terminus. The chemistry of phosphodiester hydrol…
August 31, 2026 · 37 min
Replication reports
Each automated run of a published result, with verdict, protocol, and data. Confirmations and failures alike.
CC BYReusable protocols
Versioned, tagged, machine-executable protocols that other labs can run unchanged or fork.
CC BYOpen datasets and models
Every dataset generated and every model trained, released under permissive licenses with full provenance.
CC BYThe platform itself
Registry, protocol compiler, critic, and check library, open source, so any lab can run the same loop.
APACHE 2.005 Build it with us
A public good.
A shared investment.
We are seeking two to three founding funders to build the open alternative: lasting research infrastructure that every lab can use.
$6Mover three years
Scientific and engineering staff62%
Compute15%
Cloud-lab credits10%
Operations and legal8%
Open release and community5%
Proposed allocation of the founding grant.
THE FOUNDING YEAR$2M to make
$2M to make
the first leap.
- Scientific director, one program lead, and three research engineers hired
- Platform released as open source with the first check library
- First 25 published models re-run, versioned and reported across the four verticals
- Independent critic operating on every release
- Two external labs running protocols from the registry
- Quarterly public reports on all six metrics
Year 101
Build and prove
- Open-source platform release
- 25 models replicated across energy, compute, biotech, and robotics
- Critic and check library in production
- First two partner labs on the registry
Year 202
Close the loop
- Physical validation through partner labs in each vertical
- 100+ replication reports published
- Ten external labs running protocols
- First external deployments of the platform
Year 303
Scale
- All four verticals fully funded and running
- Community-maintained check library
- Sustainability: mixed grant, partner, and service funding
- Annual open report on the state of automated research
Why philanthropic funding?
Venture capital
Venture capital closes it
A funded AI-science startup must keep its methods, data, and models proprietary to justify its valuation. The value we create is in giving those away.
Academic grants
Academia can't staff it
Building and running an automated lab is an engineering job: pipelines, registries, QC, infrastructure. Grants fund science, not the platform under it.
Philanthropy
The payoff is a public good
Reusable protocols, open replication reports, and shared infrastructure benefit every lab in a field. No single actor captures enough to pay for it.
How we measure progress
Replications published
with verdict and full record
Protocol reuse
external labs running or forking our protocols
Critic catch rate
results rejected by the critic before release
Cost per validated result
falling as checks accumulate
External deployments
labs running the platform themselves
Time to close a loop
hypothesis to published verdict
Quarterly public reports track all six metrics. Inputs → Activities → Outputs → Outcomes → Impact.
Where Tessera fits
| Capability | Corporate AI-science labs | University labs | Open-science organizations | Tessera |
|---|---|---|---|---|
| Automated experiment loop | Yes | Partial | No | Yes |
| Methods and data fully open | No | Partial | Yes | Yes |
| Independent critic before release | No | No | Partial | Yes |
| Platform others can run | No | No | Partial | Yes |
| Sustained engineering team | Yes | No | Partial | Yes |
Our proposed model alongside broad categories of research organizations; individual efforts vary. We collaborate with all three.
06 People & principles
Built by people.
For everyone.
Oleksii Vergunenko
Founder & Executive Director
Software engineer with a physics background. Built and runs a production-scale automated media pipeline whose rigor rules—versioned registry, deterministic assembly, independent critique, and failure-to-check loop—form the foundation of Tessera’s platform.
FOUNDING HIRES & ADVISORS
Scientific director
Sets the replication agenda across the four verticals
Program leads
One per vertical: energy, compute, biotech, robotics; first hired with the anchor grant
Scientific advisory board
Domain leaders from each vertical plus open-science practitioners
Governance & independence
501(c)(3) nonprofit, to be formed
The program is operated today by a Texas S corporation. A separate nonprofit is planned to carry it; until that entity exists and is recognized, no payment to us is a tax-deductible charitable gift.
Independent board, majority unaffiliated
Founder holds one seat. Board approves budgets, program choices, and any external agreement.
Scientific advisory board
Domain scientists and open-science practitioners choose replication targets and review program design.
The founder also runs a for-profit media company. It receives no priority, exclusive license, or special access. Any future agreement is non-exclusive, at arm’s length, and board-approved with the founder recused.
07 Start something that matters
The next discovery
could start
with you.
Help build a lab where every result becomes a starting point for someone else.
ov@generationix.net- Vehicle
- Operated today by a Texas S corporation; a 501(c)(3) is planned to carry the program. We will confirm in writing which entity receives any funding before it is sent.
- Anchor size
- $2M for year one; $6M over three years across two to three funders.
- Reporting
- Quarterly public reports on all six metrics, sent to funders and published.