AETHER AI · PREDATOR

Predator

AI security research, software repair, and authorized VM workflows.

Predator brings AI reasoning, persistent context, and reviewed tool use into security research and software repair. Its research engine, Crucible, works like a lockpick: it holds verified evidence steady, proposes candidate chains, checks them against the real software, and banks only supported results.

This is the public feature showcase and evidence collection. The CLI and operator engines are private, with reviewed research access.

How search works · P-BOX / HTB training · Gen3 Private Security · Crucible · MITRE chains · CVE catalog · CodePro & System-2 · Quantum Comm Ops · joint5 & IBM Fez · Results · Work with Aether · Gen3 access · Docs

How Predator chooses the next move

Keep Predator’s ordinary computer search in charge, make quantum experiments cheaper, and require evidence before trusting them with decisions. The tests described here have not demonstrated a quantum advantage.

Think of Predator choosing from a menu of possible next moves. Some moves work well together. Some require another move first. Others conflict. The goal is to pick the best allowed combination.

A QUBO is the scorecard for those yes/no choices and how they interact. Energy is the score—lower is better. The score helps rank candidates; checks against the actual software establish whether a proposed result holds. Each run records the solver it used, including any quantum simulation or hardware.

See the joint5 charts and measured limits →

P-BOX / HTB training

[P-BOX] Predator's disposable Linux VM

Give Predator a goal, confirm the VM, and follow along as it works.

P-BOX is Predator's disposable Linux pwnbox for authorized CTF and HTB missions. Predator can spawn the VM when a hosted Pwnbox is unavailable, giving it a place to work that you can watch and control.

What you can do	How it works
Confirm and start	You confirm the VM before Predator begins work within the approved mission.
Work toward longer goals	Predator works autonomously through reviewed terminal actions, reads fresh guest output, and chooses the next step within the approved mission.
Watch, take over, and resume	Watch the guest live, take control yourself, hand it back to Predator, or stop the mission.
Clean up automatically	The VM's disk is removed when you stop it or its lease expires.

The recording below includes challenge spoilers and visible lab command history.

_{P-BOX in action · Predator / GLM-5.3 · 670 recorded guest frames · Ghostlink evidence and execution roles}

About this GLM-5.3 replay

The featured GIF comes from Predator's GLM-5.3 Ghostlink mission in P-BOX. It contains 670 recorded guest frames at 1600 × 920, with blank browser-loading frames removed. The frame provenance records the retained source-frame hashes and GIF hash.

The machine guide records GLM-5.3's research and the operator's later actions. The screen recorder stopped before the final user and root challenge-file reads; those observations are documented separately.

Hack The Box: training and evidence

HTB is a test and training ground for Predator's agent and VM workflows. These two examples show autonomous VM command work and saved lab research, with completion records and execution roles in their guides.

The Satellite replay includes challenge spoilers. Both GIFs are assembled evidence replays.

Nmap course · autonomous VM work	Satellite Exploitation · recorded lab work
12/12 Academy sections completed. Predator through GLM-5.3 ran the VM commands autonomously. The owner supplied prompts and exercise questions, then entered the answers. Course record and replays →	9/9 labs completed. The certificate and account activity support completion; the replay covers four saved lab records. Certificate, lab list, and replays →

More HTB records and replay notes

Module or track	Recorded progress	Open the guide
Academy: Network Enumeration with Nmap	12/12 sections completed · October 3, 2026	Completion record and two replays
Satellite Exploitation	9/9 labs completed · October 5, 2026	Certificate, lab list, and replays
AI and ML Exploitation	1/17 labs HTB-accepted · in progress	Lab status and three replays
Machine: BlockSynergy	User and root challenge files read · October 5, 2026; owner confirmed	Machine record and sanitized replay
Machine: Ghostlink	User and root challenge files read · October 5, 2026; HTB submission unobserved	Machine record and full-screen pwnbox replay

The earlier Pwnbox console preview shows a separate reviewed guest-session check. AI/ML counts only HTB-accepted results toward completion. The machine guides distinguish Predator research from operator execution and explain what their recordings contain.

Browse all HTB modules, certificates, and replays →

Gen3 Private Security

Gen3 is the name for Predator's current research workflow and the evidence it leaves behind. It describes a method, not a measured quantum advantage.

Five scoped research records from October 2026, each with its validation and disclosure status kept separate. The Crucible research workflow is described in the next section:

Target	Public material	Status
Vercel AI SDK	Redacted local P-BOX replay	Verified local result with synthetic values; no production access.
Gemini CLI	Sanitized Crucible sprint graphic	Three private Google Cloud VRP reports under triage; no assigned severity or reward.
GitLab Duo	Private-research tactic map	Owner-controlled local validation; no vulnerability report filed.
GitLab for VS Code	VPS3 Predator P-BOX research visual	Private report filed; technical record remains private.
Anthropic SDK	Conditional AX research map	Three owner-controlled local paths from one shared boundary; no vulnerability report filed.

Vercel AI SDK: redacted local P-BOX replay

Predator's VPS3 P-BOX also supported a local security test of the Vercel AI SDK. The public replay shows the disposable Linux desktop and a verified local result. It uses synthetic values; no Vercel production account or service was accessed.

_{Five actual guest captures · seven replay frames · selective FFmpeg blur · provenance and disclosure limits}

Gemini CLI: private Crucible sprint

On October 9, 2026, Predator's manual research workflow ran a deep Gemini CLI Crucible sprint across the C1 → Q173 checkpoint span. Three separate reports were filed privately with Google Cloud VRP during the day. They remain under triage; no severity or reward has been assigned.

_{Public taxonomy only: Defense Evasion → Execution → Credential Access → Exfiltration. No reproduction details or credential material.}

Sanitized research energy

The two views show one caller-calibrated model checkpoint, with private attack-condition labels removed. Its weights are research bookkeeping; they do not independently verify a finding, measure exploit probability, establish quantum advantage, or set a Google bounty tier. Validation evidence remains in the private reports while triage is pending.

GitLab Duo: private Crucible research

Predator began an owner-controlled GitLab Duo research sprint on October 9, 2026. The graphic below is a C1 → Q48 snapshot from October 10; later manual checkpoints remain private. These labels identify research checkpoints, not executed quantum interventions or a paid CFC run. The local result maps at the MITRE ATT&CK tactic level to Execution → Credential Access → Exfiltration. This is a research map of an owned test, not a claim about an attack on another user.

Private findings: technical details have not been disclosed and no vulnerability report has been filed. No severity or bounty has been assigned. The saved evidence does not establish zero-click execution or a sandbox escape.

_{October 10 snapshot · owner-controlled validation · tactic names only · no reproduction steps, secrets, or third-party data}

GitLab for VS Code: private report filed

Predator banked an owner-controlled GitLab for VS Code research record and filed a private report. Technical details and evidence remain private while disclosure is coordinated.

Anthropic SDK: private Crucible research

On October 10, 2026, Predator banked AX001–AX003 as three conditional outcomes of one shared trust boundary in owner-controlled local fixtures. Their public MITRE ATT&CK tactic lenses are Collection, Credential Access, and Execution. These are parallel research paths, not one completed attack sequence or three independently confirmed vulnerabilities.

The saved evidence does not establish a deployed lower-trust writer route, Claude-selected tool action, real credential exposure, or deployed code execution. No vulnerability report has been filed, and no severity or bounty has been assigned. Technical reproduction details remain private.

_{Owner-controlled fixtures · parallel tactic lenses · no reproduction steps, secrets, or third-party data}

Crucible: adaptive quantum threat modeling

Think of a lockpick. Hold tension, test one pin at a time, and keep every pin that sets.

Crucible is Predator's software research and repair loop. It starts with a pinned source revision, a defined question, and fixed acceptance checks. The easiest way to picture it is a lock being picked: one hand holds tension so nothing already gained can slip, the other tests pins until one clicks, and a pin that sets stays set while the next one is worked.

_{One depth step of the loop, drawn as a lockpick. Visual by Mythos 5.1.}

The four pins illustrate a research design, not four measured workers or native checks in every run. The execution and evidence guide separates the CFC selector, an admitted exploit ratchet, and manual research. It also defines the minimum candidate record and the native effect needed for a zero-click claim. A model score, a worker proposal, and a native result are different records.

The lock, part by part

In the lock	In Crucible	What it does
Tension wrench	Master checkpoint (C)	Holds verified evidence with its provenance, signatures where present, the passes that must not be lost, and obligations still open. A manual checkpoint is not signed merely because it has a C label.
Pins	Candidate chains	Fanout proposes alternatives from a checkpoint. A proposal and its QUBO ranking do not imply that a native check ran.
The click	Native check	The software itself says whether a pin set. A chain with no native receipt did not click, whatever the model thought of it.
A set pin	HIT	Banked only with an appropriate native receipt. Misses, refutations, and inconclusive results stay in the record with their limits.
Choosing the next pin	QUBO selection and evidence gate	A modeled energy can rank a test; native outcomes determine what is supported. JEV participates only when an admitted stage records a typed call and its usage. Its decision is advisory, not an independent software result.
Working the set pin	Repair lane	Pair a supported attack observation with a repair hypothesis and regression plan. A fix counts as validated only after its own check.

What happens in one cycle

Step	What happens
Pin	Freeze the source revision, scope, evidence, and acceptance checks. This is the tension.
Fan out	Open several small candidate chains from the checkpoint. Each proposes a threat path or repair.
Model	State the target facts, candidate choices, constraints, and provenance. A bounded QUBO can rank tests; not every library chain has a populated energy profile.
Search	Record the solver actually used. A classical minimum is one valid selection method; a simulator or quantum device is a separate, explicitly evidenced run.
Validate	Run the admitted native check and retain its receipt. A proposal without that check remains a hypothesis. Record HIT, miss, refutation, or inconclusive outcome accurately.
Select and carry	Carry the supported evidence and open obligations into the next checkpoint. Pair an attack observation with a repair record; verify the repair separately.

Quantum depth

Each cycle leaves a checkpoint, and each checkpoint seeds the next intervention. A failed check is not wasted: it can reveal a missing constraint or a repair obligation that shapes the next pin.

C0 → Q1 → C1 → Q2 → C2 → Q3 → C3

Label	Meaning
C	A saved evidence checkpoint. The tension at that depth.
Q	A dependent quantum-method intervention when one actually ran. Manual C/Q shorthand can label research stages, but the label alone is not an execution receipt.
Quantum depth	How many evidenced, dependent quantum-method interventions ran in one chain. Circuit depth is a different number: the size of the circuit inside one intervention.

The ratchet is the rule that verified earlier results and negative evidence remain visible while the next candidate is checked. It does not guarantee that a later candidate will pass. In the CharLS case, Q3 used C2's remaining repair obligation to move from 16/24 to 24/24 protected test passes, retaining all 16 earlier passes.

Research roadmap →

Methods, evidence, and current limits

QUBO stands for quadratic unconstrained binary optimization. Its energy is the modeled cost to minimize; a low energy is a reason to test a candidate, not proof that it works. Only the native check on the software sets a pin.

The public CharLS and QPACK studies used quantum-circuit simulation. The joint5 pilot supplies a separate hardware record. Jev can provide structured triage only after the stage path is enabled, qualified, and deployed. A NOT_CONFIGURED field, an adapter in source, or an offline preflight is not a provider call. A participating run must record its typed decision, call count, and spend; JEV never terminates a chain on its own.

Where issued, signed custody binds research stages to source and artifacts. Manual research notes and C/Q labels are not signed receipts. Live signed acceptance, server-side stage minting, and variable-depth custody remain under validation. Each continuation needs its own scope, budget, and admission.

Gen3 seeks attributable quantum advantage over credible AI-only and classical alternatives, with matched resources and independent replication. Quantum advantage and a measured Jev contribution remain unestablished. The lockpick figure illustrates the loop's control structure; it is not a measurement, a severity, or a claim of advantage.

Where the pins come from: the ATT&CK chain library

424 source-bank chains · 18 routing buckets · October 10, 2026 V16 registry snapshot

A pin is only as good as its source and test. The library provides chain definitions, prerequisites, and references for candidate generation; a new source-grounded hypothesis can also be proposed. The 424 count is catalog coverage, not 424 tested exploits or 424 calibrated energy models. Most profile slots are explicitly unmodeled. The map shows the 15 current MITRE ATT&CK Enterprise tactics alongside Aether's AI injection, MCP, and memory-safety research areas. Predator's routing labels are maintained separately from MITRE's tactic names.

_{Open the full-size research map · GX, GLX, and AX are public case-family labels; no exploit steps are shown.}

What a chain contains	Why it helps
Connected steps	Show how a proposed path depends on earlier conditions.
Prerequisites and references	Keep the model tied to stated assumptions and cited sources.
Routing category	Organize the review around the relevant research surface.

Research mode follows chain and CVE references to develop cited hypotheses, counterexamples, and proposed extensions; it is the c-stage of a pin. Source review and appropriate checks on the software must establish observed behavior before a modeled path becomes a finding.

AX · Anthropic SDK

AX001, AX002, and AX003 mark conditional local research paths with Collection, Credential Access, and Execution tactic lenses. The paths share one underlying trust boundary and are shown separately in the Anthropic research card. The public map does not claim a sequential exploit, a deployed integration, or a vendor-confirmed finding.

CVE catalog and database

The vulnerability records the chains anchor to.

Alongside the chain library, Predator maintains a catalog of CVE records, affected products, advisories, problem types, and scoring data. CVE IDs connect those records to chain anchors so an investigation can follow the underlying evidence.

Part of the catalog	What it contributes
CVE source records	CVE JSON 5.x imports and affected-product information.
Separate enrichment	NVD, CISA KEV, and FIRST EPSS data alongside the source records.
Research review	Deep-cycle integration reads shortlisted full CVE cards before ranking and records what was reviewed.

Drive uses the approved mission's pinned evidence for investigation and validation. Catalog coverage and freshness depend on completed imports and pinned snapshots.

CodePro and System-2: how the fanout runs

Many workers, one evidence trail. When Crucible fans out, it needs a way to give several workers the right context at once and to read their answers back as one record. That is CodePro.

Part	What it does, in plain terms
CodePro fanout	In an admitted run, starts scoped workers and records each returned claim and provider receipt. A manual research turn or unavailable dispatch path is not a completed fanout.
System-2 shared IR	Aligns claims, evidence references, and open questions from workers that use the same intermediate representation.
Clean model worktrees	Source-backed runs can isolate model work in pinned Git worktrees. The run record identifies the actual checkout and revision rather than assuming isolation.
Unlimited Context (UCL)	Keeps source-backed evidence in named memory pools across sessions and models. “Unlimited” describes retrieval reach: relevant slices enter the model’s normal attention window.

Evidence-accounting update: a tested System-2 draft counts supporting sources rather than repeated statements; integration remains pending. What was tested and what remains open →

CodePro implementation links

System-2 orchestration · Native compiler · Benchmarks

Implementation links require repository access.

The private CLI around it

The private Predator CLI brings research, tools, and persistent evidence into one conversation. Choose an available Aether model, keep each run attached to its sources and approved scope, and continue from saved work. Restart checks verify the task, worktree, and memory-pool identity.

Capability	What it does
Research and Drive	Cross-reference chains and CVEs, review candidates, and run approved investigation and validation stages.
Predator RC	Follow browser and CLI activity through an encrypted relay, take over, and hand control back.
P-BOX	Spawn a disposable Linux pwnbox for a confirmed mission, with live viewing, operator control, and automatic cleanup.
Guest consoles	Work through supported Pwnbox/noVNC sessions or an owned VM in an authorized training lab.

Browser, RC, and guest-session status

Supported Pwnbox/noVNC sessions have recorded attachment, observation, reviewed input, human takeover, resume, and cleanup. An owned Ubuntu VM supplied the Academy workflow shown earlier. Other CTF, Sherlock, VM, and VPN environments need their own scope and validation.

RC is in controlled private preview. The production viewer and relay have been exercised in an owner validation; the complete sign-in, mission selection, and controller hand-back journey remains under validation. Browser and guest sessions require a fresh check after interruption.

Gen3 Quantum Comm Ops

Give each worker useful context. Verify every handoff. Focus further work on promising results.

Quantum Comm Ops is Predator's research into coordinating large agent workflows. A shared evidence bank holds the findings; a compiler prepares small, task-specific messages; receivers check those messages before using them. Adaptive Custom Fanout Controls (CFC) decide which branches need more attention next.

The programme compares ordinary delivery with an experimental quantum-search route through Nano → ATLAS → an IBM backend: first FakeFez, a simulator configured to model an IBM device, and since October 31 the real devices ibm_fez and ibm_kingston on IBM's free open plan. Compact packets select or reconstruct records from the shared bank. Useful reasoning from those records is a separate model-qualification step.

Current milestone: 400 agents, 400 messages, one IBM job, five seconds

What the mission is, in plain words. Picture a team of 400 AI worker agents. Each one needs its own short briefing before it can start its task. Quantum Comm Ops writes those 400 briefings, sends each one through an IBM quantum computer, and checks that every agent received exactly the briefing meant for it.

• Each briefing is a 96-bit message. It is cut into three small pieces and each piece becomes one tiny quantum circuit, so a mission is 1,200 circuits.
• The quantum computer runs each circuit a few times (shots) and reports the bits it read. A majority vote over those readings gives the bits of the message.
• Each message carries a check code (a 32-bit tag), like a tamper seal. The receiving agent recomputes the code with its key; if it does not match, the message is refused and that agent's pieces are re-sent.
• IBM also returns a signed receipt for the job. Predator stores it so the delivery can be re-checked later.
• "400/400 verified" means all 400 agents got a message whose check code matched, so every briefing arrived intact. "First try" means no agent needed a re-send.

In this research the 400 agents are scripted stand-ins, not 400 reasoning language models. The test measures message delivery. Whether a model can use the briefing well is tested separately (see "Evidence flows onward" below).

What happened, October 31 – November 1, 2026. The mission that previously ran on a simulator now runs on real IBM quantum hardware, and the whole thing fits in one IBM job billed 5.0 seconds of quantum-computer time. Every run was preregistered and rehearsed on the simulator first; the attempts that missed their targets or did not run are listed below and in the guide.

What we measured	Simulator (Oct 31)	IBM hardware (Oct 31 – Nov 1)
Agents whose message arrived intact on the first try (out of 400)	400, 400 and 399 on three runs	400 on all five runs: three different message sets on ibm_kingston, then two runs with a secret key (ibm_kingston, then ibm_fez)
After one re-send	400 on every run	400 on every run; no re-send was needed
IBM jobs per mission	1,200–1,203	1 (all 1,200 circuits in one request)
Quantum-computer time IBM billed	— (simulator)	5.0 seconds, down from 162 seconds in 81 jobs when the same mission first ran on ibm_fez
Wrong bits after the majority vote	1 wrong piece in three runs	0 wrong bits in 480,000 readings across the five runs; before the vote, 0.41–0.52% of single readings were wrong
Secret key check	—	With the wrong key, 0 of 400 messages pass the check (as intended); a scan of the run's files found the secret nowhere
Does packing save money? Three circuits, same device	—	Sent as one job: 2.0 s. Sent as three jobs: 6.0 s. Same result in both orders (ibm_fez)

Why it got cheaper. IBM bills by the second of quantum-computer time, and it charges much less for 1,200 small circuits in one job than for the same circuits spread over many jobs. The saving is IBM's billing for batched circuits, not a quantum speed-up: the circuits here carry ordinary, known bit patterns. The 162 → 5 s comparison also mixes different devices, days and shot counts; the clean like-for-like test is the three-circuit row above.

What did not work. A one-job run at three shots had 395 and 392 of 400 messages arrive intact on the first try, below its target of 396 (a re-send fixed the rest; five shots fixed the problem for good). A compiler setting that looked better on paper did no better on the real device and was dropped. One ibm_kingston job waited eleven hours in the queue and was cancelled. One ibm_marrakesh job was stopped by a time limit we had set too low. The 40 seconds reserved for those two jobs stay counted against the budget until IBM finalizes their bill.

Evidence flows onward. Four of the hardware-delivered briefings (one from ibm_fez, three from ibm_kingston under the secret key) were re-checked from their signed receipts, decoded, matched to the shared evidence bank, turned into a compact typed message, read by a frontier model, and the model's answer was scored by a deterministic checker: 4 of 4 correct end to end. The typed-message format itself passed 12 of 12 held-out tasks with zero formatting errors. The whole hardware programme used 254 of the 600 free seconds on IBM's open plan and about $18.65 of a $50 model budget.

The simulator milestone it builds on, three fresh 400-agent runs at three shots with a capped six-shot re-send (1,199 of 1,200 intact on the first try, 1,200 of 1,200 after re-send), remains the research baseline and is kept in full in the guide.

What comes next: feel the pins, then set them

The delivery channel is solved. The next work is on the agents, and it follows the same lockpick that Crucible already uses. In plain words:

Lockpick	What the 400-agent mission does today	What the next version adds
Hold tension	The shared evidence bank holds every verified record with its receipt	A stronger encoder picks each agent's briefing from that bank, with "this was contradicted" labels kept on the record, so fewer agents chase dead ends
Feel many pins	400 cheap agents each take one briefing (one IBM job, 5 s) and report HIT / MISS / REJECT with a short reason	Cheap models do this round; in a first survey one mid-cost model verified 12 of 12 test tasks at about 63% of the frontier model's price, and a free-tier model 7 of 12 at about 2%
Keep the pins that bind	The program's own check decides; a verdict is recorded only when it type-checks and its receipt matches	Agents that disagree are not averaged away: the split itself is recorded (a plain majority was right on 9 of 12 test tasks) and goes to a frontier reviewer
Set the binding pin	Today the loop keeps narrowing: 400 → 40 → 4, same shallow effort each round	Stop narrowing once the few real leads are known. Each surviving lead gets a deep cell: a slow "think longer" model that can loop on the problem many times, plus a frontier model, working that one lead for as long as the budget allows
Only as many hands as we have	—	Deep cells are opened only up to the machines we actually have (today: one at a time). Leads that cannot be deepened yet wait with their evidence intact; nothing is dropped for lack of a machine
Next cycle	Each round's findings feed the next round's briefings	Deep cells add records within a lead, so the bank deepens where the lock is binding instead of only getting narrower

Why this matters for a real security mission: the current loop is good at finding easy bugs and good at losing hard ones, because a cheap agent that cannot tell a hard lead is real discards it. Deepening on survivors is where the hard, high-value findings get set. Modelled from the survey above, a 400-agent round of this design costs about $4 of model calls plus 5 seconds of quantum-computer time, against about $18 with a frontier model on every agent; that is a projection, not a measured mission.

Still open, in order: the encoder comparison on a frozen dataset; whether the "think longer" model verifies anything at all (so far it has not, and until it does the deep cells use frontier models only); the narrow-then-deepen comparison at matched budget; then the repeat on ibm_kingston after a fresh calibration, packing beyond three circuits, real key management, and production dispatch. The tested superposition multiplexing idea (several messages in one circuit) ran on hardware but did not beat ordinary packing as a way to carry messages; whether it helps choose between disagreeing agents is still to be tested against a plain vote.

Status, November 1, 2026: the 400-agent delivery mission is measured on IBM hardware, including what IBM billed. No quantum advantage is claimed or demonstrated: the saving is IBM's billing for batched circuits that carry known bit patterns. Scripted agents, one synthetic message generator, two devices, one day each. The separate joint5 compiler pilot is described below.

Plain-English guide, hardware missions and open questions → · Public aggregate data and source pins

joint5 compiler and the IBM Fez result

joint5 changes how a quantum circuit is built. AQRC translates the QUBO scorecard into instructions that a simulator or device can run. joint5 is an alternative compilation path. Fewer native two-qubit gates mean fewer hardware instructions involving a pair of qubits. Answer quality is measured separately against the same scoring problem.

The IBM Fez hardware pilot

What was measured. The joint5 pilot compared ordinary and joint5 compilation on IBM Fez: one six-variable fixture, its vulnerable and fixed poles, four circuits, and 1,024 shots per circuit in one hardware job (3 seconds of billed QPU time).

Measurement	Ordinary → joint5	Change
Native two-qubit gates, each pole	144 → 103	28.5% fewer
Distance from ideal output, vulnerable pole	0.3952 → 0.3165	19.9% lower
Distance from ideal output, fixed pole	0.4039 → 0.3080	23.7% lower

Lower total-variation distance means closer agreement with the ideal output distribution. Expected energy worsened on both poles. The pilot records one acquisition without independent confirmation. Improved minimization and quantum advantage remain unestablished.

Measurements and uncertainty → · Full-precision data →

How the search and compiler scale

An October 20, 2026 synthetic sweep asked two separate questions: how close do the sampled answers get to the best score, and does joint5 save hardware instructions as the choices grow? These were simulator and compilation tests, separate from the IBM Fez acquisition above.

_{Answer quality and circuit size are different measurements. Selected sizes shown; full tables, methods, and limits are linked below.}

The proposed upgrade is simple: compare both circuit layouts, keep the smaller one, and use a suitable local simulator for quantum experiments. Ordinary computer search keeps choosing the next move until another method earns that role in a fair comparison. These are design recommendations; the sweep does not establish that the upgrades have shipped.

Scaling tests: plain-English guide and full tables → · Chart data →

Gen3 OSS: banked scientific results

Two open-source test cases, published with their data and their limits. CharLS and QPACK carry deliberately introduced faults, so the loop can be scored against a known answer. “Banked” means the result and its evidence are retained and can be checked.

CharLS Q3 repair	QPACK feedback study

Study	Recorded result	What it shows
CharLS Q3 repair	16/24 → 24/24 protected passes; eight gained, all earlier passes retained	The ratchet working as described above: a dependent repair at depth 3 closed the obligation left by depth 2, on real C++ source, without dropping a single earlier pass.
QPACK feedback study	5/8 controlled faults found with feedback, versus 2/8 without it	A second pass at the pins, informed by the first pass's native checks, found more than one pass alone in this exploratory campaign. The adaptive classical comparison was incomplete, so the quantum selector's own contribution was not isolated.

Both studies used deliberately introduced faults. A newly discovered CVE finding has not been published. Discovery, matched comparisons, and quantum advantage remain active research objectives.

Ways to work with Aether

Commission a focused repository mission, qualify a repeatable security profile, or scope an authorized live assessment.

Path	Best fit	Start here
Predator Strikes	A known CVE fix, focused discovery, or chain analysis for one agreed repository surface	Discuss a Strike →
Predator CI	Repeatable checks against a qualified repository profile and exact commit; private preview	Qualify a profile →
Predator Red Team	An authorized assessment of a live environment, with reviewed findings and remediation	Scope an assessment →

Agree on the surface, acceptance checks, and handoff before work begins. Live assessments also require approved targets, methods, timing, contacts, and stop conditions. The linked service pages list current pricing and terms; broader engineering work starts with a project inquiry.

Read the safety model → Human-approved scope and budgets bound each run. Memory, model suggestions, and quantum results supply evidence; authority comes from the approved mission. The guide distinguishes required policy from controls proven in code.

Join the community

Ask about the published studies, suggest a public demo, or share feedback in GitHub Discussions. Use issues for concrete documentation corrections.

Contributing guide · Community guidelines · Security reporting

Developers: join the research program

Help build the CLI, repair chains, verification tools, or classical/quantum comparisons. Tell us what you have built and where you want to contribute. Private research access is reviewed separately.

Apply to join →

Predator R3D and Predator BLU3

Two authorization gates for Gen3 in the private Predator CLI. Request access to Predator R3D or Predator BLU3 from the CLI. Access requests are reviewed separately from developer-program applications.

_{Predator CLI console preview · illustrative layout and counters}

Read the evidence

Start with	Then explore
P-BOX feature showcase	VM lifecycle, operator control, and recorded guest session
HTB training evidence	Courses, tracks, machines, certificates, and replay index
Crucible	The loop explained as a lockpick, with MITRE chains and the CVE catalog it draws on
CodePro and System-2	How the fanout runs, and the private CLI around it
Quantum Comm Ops	Agent mailboxes, adaptive CFC, packing and multiplexing; the 400-agent message mission on IBM hardware and aggregate data with source pins
CharLS Q3	Public data and the completed dependent repair
QPACK study	Episode data, comparisons, and missing outcomes
joint5 IBM Fez pilot	Public data, uncertainty, and the energy regression
Research roadmap	Generation goals and the requirements for promoting research into qualified profiles
Predator safety model	Required boundaries and the status of implemented controls
Unlimited Context	How it works and safety measures

_{Public overview and selected research notes · Private operator engines
© 2026 Aether AI LLC. All rights reserved. IBM is identified as a hardware provider; no affiliation or endorsement is implied. Visual sources.}