In July 2025, the Defense Advanced Research Projects Agency (DARPA) announced two programs that, read together, form a pipeline for turning indirect machine readouts alone of purported “unknown biothreat” samples into operationalizable digital protein sequences—using algorithms rather than direct observation to determine the sequences that will digitally represent those alleged threats—then “predicting” what the digitally represented biothreats supposedly do and characterizing them as threats now requiring government response, with the ultimate goal of using those digital reference sequences to accelerate medical countermeasures, potentially including vaccines.
The revelation comes as mainstream news outlets promote messaging claiming that artificial intelligence has been used “to design brand new viruses that are fully functional and can replicate in the laboratory,” indicating a broader mainstream narrative increasingly conditioning the public to regard purported AI-generated biological agents as a new and credible class of threat.
The first program, PROtein SEquencing (PROSE), was announced by DARPA’s Microsystems Technology Office on July 9, 2025.
Twenty-two days later, DARPA’s Biological Technologies Office announced the Network of Optimal Dynamic Energy Signatures (NODES) program.
PROSE and NODES are separate programs administered by separate DARPA offices.
But their respective outputs and inputs line up.
PROSE is developing technology to turn machine readings of purportedly threatening unknown protein samples into a digital amino-acid sequence.
NODES is developing computer models that begin with protein sequences and “predict their associated functions.”
DARPA’s solicitations reviewed for this report do not explicitly state that PROSE outputs will be fed into NODES.
But they describe two programs announced 22 days apart whose capabilities could operate sequentially:
purported unknown biothreat sample → machine signals → algorithmically called digital sequence → computationally predicted function → threat characterization → medical countermeasures.
It all ends with medical interventions and government “threat” response.
Follow us on Instagram @realjonfleetwood & Twitter/X @JonMFleetwood.
If you value this reporting, consider upgrading to a paid subscription.
For advertising & sponsorship opportunities, contact us by clicking below.
DARPA Says It Needs New Technology to ‘Read’ Previously ‘Unknown’ Biothreats
PROSE begins with an extraordinary premise.
DARPA says the Department of Defense cannot rapidly identify certain purported biological threats, including threats that are “unknown.”
According to the agency’s BAA (found here):
“The Department of Defense (DoD) lacks the ability to rapidly detect and identify unknown biothreats, such as natural and engineered protein-based biotoxins.”
DARPA then claims artificial intelligence and synthetic biology are creating a new class of purported biological threats:
“New capabilities in synthetic biology and artificial intelligence (AI) enable the rapid design and production of novel protein-based biothreats.”
The agency says freely available AI protein-design tools can:
“generate thousands of re-engineered toxin molecules with similar structure, and likely function, as the parental toxin, but with such radically different sequences that they are unidentifiable as threats with current technology.”
DARPA says PROSE will:
“develop a high throughput sequencing technology to fill this critical gap in the DoD’s threat response capabilities.”
But the purported threats DARPA wants PROSE to “read” do not necessarily have known genomes, or even genetic material.
According to the BAA:
“While current methods can rapidly identify threats with known genomes, we lack the capability to read biothreats that do not have genetic material, such as biological toxins, or threats with currently unreadable modifications that augment function.”
DARPA nevertheless says PROSE will:
“enable the identification and characterization of these threats”
and provide technology for:
“threat/pathogen identification, warfighter health, in-field decision-making, and intelligence gathering.”
The program description then identifies what PROSE is supposed to “read”:
“PROSE will demonstrate molecular readers that can accurately read a broad range of amino acids and post-translational modifications (i.e., letters) in sequence for unknown protein samples.”
DARPA says those readers must handle sequences of at least 300 “letters,” at claimed accuracy of at least 99%, and at a throughput of at least 10¹⁰ letters per day.
And DARPA anticipates that synthetic biology will dramatically expand what can count as a “letter”:
“emerging techniques in synthetic biology are expected to increase the number of possible letters exponentially over the next decade.”
PROSE readers must therefore demonstrate:
“scalability across a broad range of chemical complexity.”
DARPA says accomplishing this will require new protein read elements, microsystems, translocation approaches, measurement technologies, and algorithms.
But what exactly will these new molecular readers directly observe?
PROSE Does Not Directly Observe the Sequence It Reports
This is where the distinction between measurement and interpretation becomes critical.
PROSE does not directly observe a string of amino-acid letters inside the purported biothreat sample.
The proposed machines generate signals.
Software then determines what those signals supposedly mean.
DARPA explicitly requires:
“algorithms for letter calling capable of translating raw signal data into a specific amino acid sequence.”
DARPA calls this process “letter calling.”
The agency describes PROSE as demonstrating molecular readers for purported “unknown protein samples,” followed by:
“advanced algorithms that translate signals from the integrated system into letter calls.”
The distinction is fundamental.
The machine signal is the measurement.
The amino-acid letter is the algorithm’s interpretation of that measurement.
And the sequence is assembled from those interpreted letters.
Electrical Current & Fluorescence Become Biological ‘Letters’
DARPA’s solicitation provides examples of what these indirect measurements can look like.
Discussing existing nanopore and optical technologies, DARPA explains that chemical information can be transduced into:
“a change in current in a circuit, or a fluorescence signal read by a camera.”
Those signals are not amino acids.
They are electrical or optical outputs interpreted as corresponding to amino acids.
PROSE is intended to push this principle considerably further by developing “novel measurement modalities” capable of distinguishing a much larger universe of purported protein “letters.”
DARPA explicitly describes the final computational step as:
“advanced algorithms to translate integrated system signals into a called sequence of letters.”
The proposed chain is therefore:
purported “biothreat” protein sample
↓
physical interaction with a reader
↓
machine signal
↓
algorithm
↓
amino-acid “letter call”
↓
quality/accuracy score
↓
digital protein sequence
DARPA even requires performers to document the computational models responsible for:
“translating raw signals into amino acid calls”
including their architecture, training data, and performance metrics.
In other words, the digital sequence is not the raw machine observation.
It is the product of an interpretive computational process applied to the raw machine output.
The Digital Sequence Can Then Stand for the Alleged Biothreat
That distinction becomes considerably more consequential because PROSE is specifically intended to operate where an existing reference sequence may not exist.
DARPA starts with something it characterizes as an unknown biothreat.
The sample generates machine signals.
Algorithms translate those signals into “letter calls.”
Those calls become a specific digital amino-acid sequence.
For downstream systems and decision-makers that do not independently possess and examine the originating physical sample, the digital sequence can therefore become the operationalizable representation of the purported threat.
And DARPA explicitly intends PROSE for consequential applications.
The agency lists, as mentioned above:
“threat/pathogen identification, warfighter health, in-field decision-making, and intelligence gathering.”
The digital representation of the sequence is therefore not merely an academic description stored on a researcher’s computer.
DARPA is developing PROSE so its outputs can be operationalized.
And 22 days after announcing PROSE, DARPA announced another program built to start with protein sequences.
Twenty-Two Days Later, DARPA Announced NODES
NODES begins farther downstream.
It does not need to begin with the original purported biothreat sample.
Its stated input is a protein sequence.
DARPA says in its NODES BAA (found here) that the project will:
“develop computational models that will input protein sequences and predict their associated functions.”
That means the output datatype PROSE is being developed to produce is precisely the datatype NODES is being developed to consume.
PROSE: machine signals → digital protein sequence
NODES: digital protein sequence → predicted function
Again, DARPA’s solicitations do not establish that the agencies will specifically feed PROSE-generated sequences directly into NODES.
But the technical compatibility is unmistakable.
The two programs were announced just 22 days apart.
NODES Starts With Protein Sequences & ‘Predicts Their Associated Functions’
Where PROSE is designed to produce a digital protein sequence from machine signals, NODES begins with the protein sequence.
DARPA states:
“The Defense Advanced Research Projects Agency (DARPA) is soliciting proposals to develop computational models that will input protein sequences and predict their associated functions based on protein movements (dynamics) observed during folding, protein binding, and/or allosteric interactions.”
DARPA says those predictions will then be tested under conditions determined by the agency:
“The ability to predict protein function will also be tested across a range of scenarios defined by DARPA.”
The models will not merely simulate molecular dynamics.
According to DARPA, performers will be engaged in:
“simulating, learning, and generating molecular dynamics”
and will be required to create:
“an Application Programming Interface (API) for general usage by the community”
along with:
“appropriate guardrails to ensure the safety of both the models and the interpretation of the predicted protein functions.”
DARPA then explicitly connects those computational predictions to de novo sequences, unknown agents, threat characterization, and medical countermeasures:
“Together, these efforts will bolster the Department of Defense’s (DoD) ability to probe the limitless space of de novo protein sequences, provide a tool to expedite threat characterization when the nation or warfighters are introduced to an unknown agent, and shorten the time to developing Medical Countermeasures (MCMs).”
DARPA also says NODES will:
“support biomedical research by providing expedited ways to understand infectious, protect crops, develop new pharmaceuticals, and elucidate mechanism of disease.”
The sequence DARPA describes is therefore explicit:
protein sequence → computational prediction → interpretation of predicted function → “unknown agent” → “threat characterization” → Medical Countermeasures.
DARPA Calls Protein-Function Determination a ‘Universal Design Tool for Controlling Biology’
DARPA’s description of why it wants this capability is even more striking.
The agency begins:
“Proteins are essential building blocks of life, playing a crucial role in countless biological processes.”
DARPA says their universality makes them useful for medicine:
“The universality of proteins across lifeforms makes them a powerful tool in medicine, where their tunable properties can be harnessed to develop life-saving treatments.”
But DARPA immediately presents the other side:
“Conversely, proteins are also implicated in toxic functions and their misuse can prove detrimental to life.”
It then states what determining protein functions could ultimately provide:
“The determination of protein functions, therefore, offers the opportunity to develop a universal design tool for controlling biology in areas of early preparedness, rapid response, and derisking applications.”
And DARPA explicitly connects that capability to surveillance and national security:
“Pushing the envelope in these areas is crucial to strengthen our bio surveillance, security, and attribution efforts.”
So DARPA is not merely proposing software that assigns academic descriptions to protein sequences.
The agency itself describes protein-function determination as an opportunity to develop a “universal design tool for controlling biology” for “early preparedness” and “rapid response,” while strengthening “bio surveillance, security, and attribution efforts.”
Deep Learning Will ‘Generalize Predictions’ & ‘Rapidly Infer’ From a Given Sequence
And NODES’ computational predictions do not end with its initial models.
After creating what DARPA calls a:
“library of molecular movements”
the agency says:
“deep learning methods are sought to generalize predictions and rapidly infer an ensemble of structures with correct thermodynamics, energy, and force field-based weights for a given sequence.”
That final phrase returns NODES to its fundamental input:
“a given sequence.”
From that sequence, DARPA proposes computational models that “predict” function, systems that “simulate,” “learn,” and “generate” molecular dynamics, and deep-learning methods that “generalize predictions” and “rapidly infer” structures.
Those computational outputs are then intended to help DARPA “probe the limitless space of de novo protein sequences,” characterize an “unknown agent,” strengthen “bio surveillance, security, and attribution,” and shorten the time to Medical Countermeasures.
Read directly after PROSE, the potential handoff is stark:
PROSE: machine signals → algorithmic “letter calls” → digital protein sequence.
NODES: digital protein sequence → “generalize predictions” → “rapidly infer” → “predict their associated functions” → “threat characterization” → Medical Countermeasures.
PROSE & NODES Form Two Halves of the Same Potential Pipeline
DARPA announced PROSE on July 9, 2025.
It announced NODES 22 days later.
PROSE seeks to solve one problem: How can machine signals from an unknown sample be converted into a protein sequence when no reference sequence exists?
NODES begins with the resulting datatype and asks another: Given a protein sequence, how can computation predict what it does?
The first produces the digital representation.
The second operates on digital representations.
Both are explicitly directed toward purported unknown or novel biological threats.
And both are connected to consequential government response capabilities.
DARPA has not stated in the solicitations reviewed here that PROSE and NODES constitute one formally integrated program.
But read together, their respective inputs and outputs reveal a technically compatible pipeline:
PURPORTED ‘UNKNOWN’ BIOTHREAT
↓
PROSE
INDIRECT MACHINE READOUT
↓
algorithmic “letter calling”
↓
OPERATIONALIZABLE DIGITAL SEQUENCE
↓
NODES
digital sequence as input
↓
computational modeling / AI
↓
“PREDICT THEIR ASSOCIATED FUNCTIONS”
↓
“THREAT CHARACTERIZATION”
↓
GOVERNMENT RESPONSE
↓
MEDICAL COUNTERMEASURES
Bottom Line
DARPA is developing systems in which purported “unknown biothreats” can be assigned digital sequences and alleged predicted functions that feed into “threat characterization,” government response, and medical countermeasures—without direct observation of either the sequence or function being claimed.
PROSE turns indirect machine signals into algorithmically called sequences. NODES takes protein sequences and uses deep learning to “generalize predictions,” “rapidly infer” structures, and “predict their associated functions.”
The concern is what happens downstream.
How much confidence and government authority should be invested in an alleged threat whose digital identity is derived from machine signals and whose alleged function is computationally predicted?
What evidentiary threshold turns those inferences into “threat characterization”?
And when does that characterization become sufficient to drive government response or medical countermeasures?
The ultimate question is simple:
When does an algorithmic inference about an “unknown biothreat” become an operationalizable government fact—and what happens when the government acts on it?
Follow us on Instagram @realjonfleetwood & Twitter/X @JonMFleetwood.
If you value this reporting, consider upgrading to a paid subscription.
For advertising & sponsorship opportunities, contact us by clicking below.
If I tried to use AI this way I'd be automatically reported to the government for potential bioterrorism activity..
Why is there not a hard stop on all of this nonsense after the Covid fiasco! 🛑