165: Chapter 165 Making the World Queue

August 31, 2022, 06:01.

Room 402, Building 17, Zijing Apartment.

Jiang Lin sat in front of his desk and lowered his head to look at his right hand.

His skin was clean, his knuckles smooth, and his palm held only the thin calluses proper for an eighteen-year-old youth. The low temperatures, wind-blown sand, and cracks left by forty years of maintenance work on the Wasteland had all vanished, nor could he feel the neural blockage that had gradually worsened in his old age when he made a fist.

His body had returned completely intact.

The desk lamp was still on.

The laptop, military training cap, folded camouflage jacket, and the anti-chafing tape opened and left by the bed last night all remained where they had been a minute ago.

Jiang Lin sat for over ten seconds, waiting for his bodily sensations to stabilize before reaching for the portable hard drive protective case on the left side of the desk.

A white label was stuck to the outer shell of the protective case.

[J-11/Reality Return/Sealed]

He checked the latches, tamper-evident seals, and internal shockproof pads, confirming no signs of displacement on the casing before connecting the data cable to the offline laptop.

The terminal opened.

Jiang Lin typed in the first line of command.

[sha25J11_manifest.sha256]

Enter.

The cooling fan began to spin, and verification results scrolled line by line across the screen.

[G01C_HeatRouter_FieldPack_v1.0.tar.gz OK]

[MPS_HeatRouter_Reality_v2.0.tar.gz OK]

[HeatRouter_Node_Prototype_v1.0.tar.gz OK]

[PMCU17_ShadowBus_Handshake_v1.0.tar.gz OK]

[MPS_FaultLedger_Reality_v0.9.tar.gz OK]

[OuterRing_FaultLog_Decoder_v1.2.tar.gz OK]

[FDSO_207Map_v0.4.tar.gz OK]

[TM7_InnerRing_Access_Prerequisites_v0.1.tar.gz OK]

[Thermal_Load_Migration_Criterion_Draft.tar.gz OK]

[Stellar Tide_Preparatory_Project_Index_v0.1.tar.gz OK]

Nine result packages and one catastrophe index all passed verification.

Jiang Lin plugged in the second cold backup drive.

Copy.

Re-check.

Disconnect.

He then wrote it into the local Git index, generating a read-only image.

At 6:18, the three copies completed cross-verification.

Jiang Lin unplugged the original hard drive, locked it back into the protective case, and created a new file on the offline laptop.

[J11_Reality_Reentry_Triage.md]

The file contained only three columns.

[Public Collaboration Layer]

[Controlled Review Layer]

[Absolute Archival Layer]

The Public Collaboration Layer was left empty for now.

Materials, machining precision, heat dissipation conditions, and operating environments in reality differed from those at the Outpost.

Even for results that had been downgraded and rewritten, they had to be reproduced on a physical prototype first; at this stage, nothing met the criteria for public disclosure.

Jiang Lin moved the G-01C Heat Flow Router Field Pack, MPS-HeatRouter, Computing Node Prototype, MPS-FaultLedger, and Thermal Load Migration Criteria into the Controlled Review Layer.

These files could be handed over to the Low Entropy Workshop for internal reproduction, but every transfer required generating an independent copy, stripping away all Wasteland designations, material sources, and long-term experimental records, leaving only design objectives, boundary conditions, and test interfaces that the real-world team could verify.

The remaining five files were placed into the Absolute Archival Layer.

PMCU-17 Shadow Maintenance Bus.

Outer Ring Fault Log Decoder.

FDSO-2076A Fragment Map.

TM-7 Inner Ring Access Prerequisites.

Stellar Tide Preparatory Engineering Index.

They would not enter Low-Entropy, would not enter Tsinghua University, nor would they enter any networked device.

Only after completing the triage did Jiang Lin turn on his phone.

The notification bar was stacked with military training group messages, Mathematical Sciences Center system notifications, Low Entropy Workshop internal daily reports, and a batch of email reminders.

He first clicked open the operation log synchronized at dawn by the Research Support Unit.

[Jiang Lin Research Support Unit (Temporary) / Operation Log / 2022-08-30]

[Archived Mailbox Total Inbound: 312]

[Category A: Technical Question List] 9

[Category B: Seminars and Reading Groups] 46

[Category C: Media Interviews] 83

[Category D: Commercial Cooperation and Investment Offers] 41

[Category E: Congratulations, Private Contacts, and Vague Offers] 133

The processing records were also concise.

[Category C transferred to university publicity desk; no private interviews accepted, no personal life materials provided.]

[Category D involving Low Entropy Workshop transferred to public business email; direct refusal for those seeking commercial endorsement using PFR/Marton.]

[Category B scheduled and organized by academic secretary, not routed to personal instant messaging.]

[Category A compressed into question list, submitted for review during fixed windows. All emails directly requesting Jiang Lin himself to solve problems returned.]

Attached at the very bottom of the log was an item pending confirmation.

[First batch of Category A-1 candidates: three items, expected to be submitted in today's evening fixed window.]

After reading it, Jiang Lin did not open the attachment.

He created a controlled task in the internal system of Low Entropy Workshop, writing only the file name and processing requirements, without uploading any data.

[Task: G-01C Long-Period Thermal Management Reproduction Preparation]

[Status: Awaiting Low-Entropy to complete isolated reproduction environment and first version Bill of Materials]

[Restrictions: No publicity requests using this task, no performance commitments to Hengtai, no bypassing physical prototype verification.]

At 6:34, the task was saved.

Jiang Lin turned off his phone, washed up, changed into his military training uniform, and went downstairs to eat breakfast.

At 7:26, he walked into Zijing Playground.

The Qiuzhen Class of 2022 had already begun assembling.

At 9:58 AM.

Third floor of the Mathematical Sciences Center, Tsinghua University.

In an office that had been cleared out not long ago, the air conditioner was set very low, blowing cold air noisily towards the floor.

This was not the personal office assigned to Jiang Lin by the center when he reported on August 16.

That office was located at the end of the second-floor hallway, about twenty square meters in size.

Inside was only a large solid wood desk and a whiteboard occupying an entire wall—quiet and private enough.

As an independent research space for a freshman who had just entered the university, it was almost luxurious, even breaking the center's years-old precedent.

But right after the "PFR/Marton Public Version" paper went online, this originally spacious office was clearly no longer sufficient.

The issue was not the small physical floor area.

If in this small domain it was only Jiang Lin alone reading papers, deriving formulas, writing proofs, and conducting offline data verifications, that office would be completely enough—it could even fit a folding bed.

What truly made it cramped was that it simply could not handle the massive volume of affairs surging in furiously from all over the world at this moment.

The classification and archiving of international emails, follow-up feedback from various seminars, massive lists of questions regarding formal verification systems, overt and covert appointments from top visiting scholars, pervasive media filtering, commercial cooperation invites requiring physical isolation, technical liaisons crossing departments and even national boundaries...

These matters carrying various demands and interests could not be piled like waste paper on Jiang Lin's own workbench.

Even less should they—and absolutely were they not permitted to—be placed in the same physical space as his personal core research materials.

Therefore, the senior leadership of the Mathematical Sciences Center made a prompt decision to vacate this higher-spec office on the third floor, originally used for small project coordination, to serve as the temporary workspace for the Jiang Lin Research Support Unit.

There hadn't been time to hang a formal bronze plaque at the door.

Only a printed A4 sheet of paper was neatly taped to the white wall beside the door with transparent tape.

[Jiang Lin Research Support Unit (Temporary)]

The room was a full size larger than the one on the second floor.

The original sofa inside had been removed, replaced with three workstations.

The three workstations belonged to the academic secretary, executive assistant, and technical liaison respectively.

The executive assistant handled media, visits, and commercial processing; the academic secretary organized seminars, version feedback, and academic letter indexes; the technical liaison was responsible for inspecting technical question definitions, attachments, data formats, and interface conditions.

The spot by the window was reserved for the academic secretary assigned by the Mathematical Sciences Center.

At the workstation near the side of the door sat the executive assistant responsible for liaising with intricate internal and external university affairs.

A narrow conference table was added to the center of the room, and the dual-screen workstation against the wall had different account permissions from the other two computers.

A long, narrow conference table was temporarily squeezed right in the middle of the room.

Meanwhile, at the innermost position against the wall with the best privacy sat the technical liaison's high-performance dual-screen computer desktop.

On the academic secretary's desk, a stack of seminar application forms printed during overtime last night and key summaries of overseas academic emails had already piled up, covered in dense annotations made with highlighters of various colors.

Brow furrowed, the executive assistant was mechanically and rapidly batch-stripping media invitations—which used academic exchange as a guise but actually sought to probe for intelligence—from the internal email system and dumping them into a designated interception folder.

The technical liaison officer sat at the innermost end, staring at the two screens in front of him.

Of the more than three hundred emails that entered the inbox last night, more than half had already been filtered out.

At 10:06 AM, the technical liaison opened the first Category A-1 candidate email.

The email came from a young teacher in Tsinghua's Department of Computer Science; the body text was very short, but the attachment directory was listed very clearly.

[Problem Background: Busy Beaver Challenge / 5-state 2-symbol Turing machines]

[Current Target: Three types of non-halting deciders: Cyclers, Translated Cyclers, and Backward Reasoning.]

[Request Content: Review whether the proof certificates output by the three types of deciders can share a minimal trusted verification interface; if not, clarify at which layer the semantic conflict occurs.]

[Known Difficulties: The three types of certificates rely respectively on full configuration repetition, translated configuration repetition, and backward unreachable sets; the internal team draft cannot yet unify state objects.]

[Boundary Statement: No request to provide the BB(5) value, no request to decide holdout machines, no request for Jiang Lin to join community signatures.]

[Attachment Materials: Minimal implementations of the three types of deciders, unpublished proof certificate drafts, passed test cases, unified failed test cases, and public community links.]

The technical liaison checked item by item.

Problem definition clear.

Proof target exists.

Attachments reproducible.

Involves no commercial or confidential matters.

He added a record to the candidate pool.

[Category A-1 Candidate / Formal Verification / Computability / Materials Complete / Requires Jiang Lin's Judgment]

The second email came from another research group in the Department of Computer Science, accompanied by a joint statement from a domestic reconfigurable computing team.

[part:gemini-3.5-flash-lite]

[Problem Background: Dedekind number DBoolean functions]

[Current Scheme: The first round of equivalence class merging has been completed according to the variable permutation orbits; the core counting tasks are mapped to the FPGA array.]

[Existing Exception: After the secondary partitioning, some oversized orbits exhibit load imbalance, cross-block duplicate verification, and hash summary expansion.]

[Request Content: Review whether the secondary task splitting has omitted canonical representatives or stabilizer information; evaluate whether the existing verification topology can be re-blocked while retaining the capability for independent re-verification.]

[Boundary Declaration: No request to provide the complete algorithm, no request to participate in computing power consumption, and no request to endorse the final D(9) result.]

[Attachment Materials: Task splitting logic, orbit scale distribution, stabilizer field description, hash verification draft, and small-scale reduction samples.]

The technical liaison did not evaluate the other party's mathematical scheme, but only checked whether the attachments could support the request.

The fifty pages of materials contained the task topology and small-scale samples, but lacked the actual runtime statistics of various task blocks.

The questions raised in the email could be submitted for review, but whether Jiang Lin could make a judgment based on them still required his personal decision.

A second record was added to the candidate pool.

[Candidate A-1 / Combinatorial Counting / Verifiable Computation / Missing Load Statistics / Requiring Jiang Lin's Judgment]

The third email was forwarded through Professor Lu Zhixing, and the sender came from a quantum information team at USTC.

In the forwarding note, Professor Lu Zhixing confirmed two things: the source of the materials was authentic, and the other party accepted all communication rules of the Research Support Unit.

As for whether the question was worth Jiang Lin handling, it was still filtered according to the normal process.

[Problem Background: Syndrome measurement and detection events in superconducting quantum circuit surface code experiments]

[Current Problem: The basic decoding model already includes local spatial correlation terms, but between different calibration windows, the residual correlation graph exhibits unstable drift.]

[Request Content: Judge whether the drift exceeds the range of statistical fluctuations; if it exceeds, further distinguish between stable correlation structures and spurious correlations generated by readout links and clock alignment.]

[Boundary Declaration: No request to locate specific devices, no tracing of error mechanisms at the process level, no request for decoder optimization; no provision of real hardware numbers, device processes, microwave control parameters, and complete physical topologies, providing only anonymized channel adjacency relationships and readout line groupings.]

[Attachment Materials: Two million measurement records, including timestamps, measurement rounds, anonymous channel numbers, boolean syndrome readings, detection event tags, and basic statistical summaries.]

The technical liaison checked up to this point and paused for a moment.

The problem definition in the email was clear and the data volume was large enough, but the attachment list lacked spatial adjacency relationships, stabilizer types, readout groupings, and calibration window markers.

The other party wanted to judge the correlation structure, but did not submit all the metadata required to construct the correlation graph.

He did not complete the problem for Jiang Lin, but only noted the missing items in the candidate pool.

[Candidate A-1 / Quantum Information / Experimental Error Structure / Data Entity Temporarily Suspended / Requiring Jiang Lin's Judgment]

At 11:22 AM, the three emails were compressed into a two-page summary.

The academic secretary reviewed the sources and citations, the administrative assistant checked the communication boundaries, and the technical liaison confirmed the attachment index and problem format.

The three of them respectively signed their process numbers, and the summary immediately entered the pending transmission queue.

...

At 3:00 PM, Zijing Playground.

The instructor blew the rest whistle, and the formation was dismissed on the spot.

Zhao Chengyu sat down by the shade of a tree, placed his military training cap on his knees, unscrewed his water bottle, and gulped down a few mouthfuls.

"I retract part of my criticism of the training schedule from yesterday."

Jiang Lin sat down beside him: "Which part?"

"The forty-five-minute mandatory shutdown and ten minutes of heat dissipation." Zhao Chengyu fanned himself with the brim of his cap, "At least thermal protection was written into this scheduling system."

"The temperature is high today, so the afternoon training session was shortened by five minutes."

"And dynamic frequency scaling."

Jiang Lin smiled slightly and lowered his head to drink water.

The phone in the compartment had just vibrated once, but he did not take it out.

External information during training was all postponed, and emergency matters were conveyed through the official channels of Qiuzhen College and the Mathematical Sciences Center. Once the rules were established, they needed to be executed stably.

The rest whistle ended, and the squad reassembled.

At 8:10 PM.

After taking a shower and changing into a clean pure cotton T-shirt, Jiang Lin sat at the desk and opened the networked laptop responsible for external affairs.

The summary from the Research Support Unit had arrived in his inbox.

[Candidate Summary of Class A Problems / First Batch / Please Confirm Processing Level]

The body text had only one explanatory paragraph.

[A total of three items in this batch. The Support Unit only completed source confirmation, boundary checking, material indexing, and problem format compression, without making judgments on the technical content. Please confirm whether to return, supplement materials, or enter my fixed processing window.]

Jiang Lin first established an archiving directory.

[2022al_Problems]

Then he read them item by item.

The first item, Busy Beaver 5.

In the territory of computer science, Busy Beaver 5 is an ancient mystery born in the 1960s.

Its past life originated from a pure question among mathematicians.

If a program is only allowed to possess very few internal states, how complex a behavior can it exhibit before entering a halting state?

In order to find the answer, they cast their gaze upon the theoretical ancestor of modern computers—the Turing machine.

They attempted to find the champion machines that run for the most steps and are the busiest under the restriction of varying numbers of states.

The task of finding the champions of one to four states has been successively determined and verified over the past few decades.

However, when the number of states reached five, the number of machines expanded drastically; what was even more difficult was that no matter how long a machine ran, it could not automatically prove that it would never halt.

This is its present life: BB(5) is no longer a pen-and-paper arithmetic problem, but has evolved into a large-scale project that requires the collaboration of global computing power, heuristic algorithms, and formal verification tools to advance together.

Its underlying rules are actually exceptionally simple.

Five states, two symbols, blank tape.

If all those abstract definitions are stripped away, it very much resembles an extreme competition specially designed for mechanical wage earners.

The contestant is a robot with a very small brain capacity.

In front of it lies a grid tape extending infinitely to the left and right, where each cell can only be 0 or 1. Every step the robot takes, it must look down at the current cell and complete three things according to the pre-written "Working Regulations".

Change the number under its feet to 0 or 1.

Move one grid to the left or right.

Switch itself to the next working state.

Five states mean that it only has five internal states, A, B, C, D, and E, when working. The regulations also allow it to enter an additional halt state under certain circumstances. Once entered, the entire machine completely stops.

What is discussed in the story is the step version of the Busy Beaver.

The goal of the competition is very simple: among all five-state robots that can eventually halt, find the one that works for the most steps.

Stopping early means the score is not high enough.

Never stopping means there is no score either.

The real trouble is that people often cannot tell which category a certain machine belongs to.

If it stops after running one hundred steps, one only needs to completely replay those one hundred steps, and the result can be verified.

But if it has already run one trillion steps and is still moving back and forth on the tape, people can only confirm that it did not stop in the first one trillion steps. It might enter the halt state in the very next step, or it may have long fallen into an extremely covert loop, repeating the exact same set of work forever.

Continuing to wait cannot solve this problem.

The tape has no end, and time has no preset upper limit. For any arbitrary program, there is no universal tool that can predict in advance whether it will eventually halt or not. This is the boundary drawn by the halting problem.

After all, the types of five-state machines are limited, and people can still choose a more difficult path: targeting the behavioral characteristics of different machines, searching respectively for evidence that proves they will never halt, and then eliminating these machines batch by batch.

The three types of deciders sent by the Department of Computer Science of Tsinghua University undertake precisely this kind of work.

The first kind, Cycler.

What it searches for are machines spinning in place.

If the robot returns to the same position and the same working state at two different points in time, and the 0s and 1s on the entire tape are also completely identical to the previous time, then all conditions at this moment have been restored. What happens next will simply replay that intermediate process all over again.

Like a stuck cassette tape.

Loop once, loop ten thousand times, the result will never change again.

The second kind, Translated Cycler.

This type of machine does not return to the original place, but works like a rolling seal.

It leaves a pattern somewhere, moves a few squares to the right, and leaves the same pattern in the same state; then continues to move to the right and repeats once more. As long as this translational structure can persist, and the old traces left behind in the past will never block it again, it will work all the way to infinity on the tape.

The third kind, Backward Reasoning.

Instead of chasing after the robot running forward, it stands at the endpoint of "halt" and works its way backward to find the path.

To enter the halt state, what positions, what working states, and what must be written underfoot might the previous step have been in? Going one step further back, what other circumstances could there be?

If all possible paths leading to the endpoint are traced forward layer by layer, and it can eventually be proved that none of them connect to the initial tape consisting entirely of 0s, it means that no matter how this machine operates starting from the specified starting point, it can never reach the halt state.

All three tools are catching machines that "never get off work", but the evidence they hand back is completely different.

The Cycler hands over an endless loop connected head-to-tail.

The Translated Cycler hands over a seal constantly rolling into the distance.

The Backward Reasoning hands over a route map traced forward from the endpoint, which is eventually completely disconnected from the starting point.

Each tool can be equipped with a set of dedicated verification programs.

However, the dedicated verification program itself may be equally complex. When it claims that a certain piece of evidence is valid, the outside world still needs to trust that its code has not omitted conditions, nor quietly accepted erroneous results during millions of judgments.

The Department of Computer Science hopes to establish yet another audit window that is sufficiently small.

This window is not responsible for searching for evidence, but only for checking evidence. Regardless of which tool has caught a never-halting machine beforehand, the materials must eventually be sent here, and the final confirmation is completed by a set of underlying programs that are sufficiently small in scale and capable of being independently audited.

The problem consequently fell before Jiang Lin.

Can a loop, a rolling seal, and a disconnected route map really be handed over to the same audit window?

What facts does this window need to inspect personally? And what conditions are merely claimed to hold true by the evidence generation program?

As long as this boundary remains ambiguous, so-called unified verification might just be rewriting the conclusions given by the three sets of complex programs all over again.

Jiang Lin did not reply directly on the summary.

Based solely on three minimal implementations and a few failed samples, it was still impossible to determine whether the three types of evidence should be forcibly compressed into the same format, or whether they only needed to share the bottom-most machine rules and audit boundaries.

He wrote his entry comments in the log.

[Initial Screening: The problem definition is clear, and the existing failed samples are sufficient to initiate independent reproduction.]

[Pending Verification: Whether the three types of decision procedures can be handed over to the same small verifier requires first reproducing the evidence they submitted, the verification methods, and the parts that must be trusted.]

[Current Action: Reproduce only, do not modify, do not submit conclusions.]

The second item, Dedekind D(9).

This problem is more like a switch game whose scale has spiraled out of control.

There are nine switches and a lamp on the table.

Now you can design the lighting rules arbitrarily. For example, turn on number one and number two at the same time, and the lamp lights up; or turn on number three alone, and the lamp lights up; or numbers four, five, and six must be turned on simultaneously for the lamp to light up.

There is only one restriction on the rules.

Once the lamp lights up, continuing to turn on more switches must not cause it to go out instead.

What the Dedekind number calculates is not how many switch combinations there are in total for nine switches.

What it wants to count is: how many sets of lighting rules satisfying this restriction can be designed in total.

Every single set of rules must answer whether the lamp lights up under various open and close states of the nine switches. Even if a certain two sets of rules differ in their answers on only one extremely obscure switch combination, they are still counted as two sets of rules.

When there are only two or three switches, you can still slowly list them out with paper and pen.

When increased to eight, the total number of rules has already reached:

[56130437228687557907788]

A twenty-three-digit number.

Increasing it to nine again, generating and checking set by set no longer has realistic feasibility.

To compress the task, mathematicians will not save a whole thick manual of lighting instructions, but will only record the minimum combinations in each set of rules that are "just enough to make the lamp light up".

Assuming that turning on number one and number two simultaneously is already able to light it up, then turning on number one, number two, and number three together will inevitably also light it up. The latter piece of information can be directly deduced from the former without needing to be saved repeatedly.

What is ultimately left behind is a batch of minimum lighting conditions where no one completely contains another.

This batch of conditions is called an antichain in combinatorial mathematics.

Even after completing this step, the quantity remains massive.

The second layer of compression comes from the symmetry of the switch numbers.

If a set of rules swaps number one and number two, it will become another set of rules. Thus, the two are actually the same template structurally, only with different switch labels. The computer can process only one template and then multiply it by how many kinds of numbering it can swap out.

The email calls all renamed versions of the same template an "orbit".

The opposing team has already completed this step, and the real difficulty occurs after sending the tasks into the FPGA array.

You can think of those orbits as cargo boxes categorized in advance.

Some boxes are very light, and a single worker can finish counting them in a few minutes; some boxes are shockingly heavy, and a single worker might not even be able to finish processing them in several months.

If one insists on whole-box allocation, the classification relationship is the clearest, but a few FPGAs will be fully loaded for a long time, while other chips will be left idle early on.

If the giant cargo boxes are split apart, all chips can be assigned close workloads, but data that originally belonged to the same category will scatter to different nodes. The system must additionally record which box each piece comes from, whether it has been calculated elsewhere, and finally how to piece it back together without omission or duplication.

The more shattered the split, the more uniform the calculation.

The more shattered the split, the heavier the repeated verification and node communication as well.

What the team is stuck on right now is this balance point that cannot be decided by intuition.

Existing attachments show that the extra overhead indeed appeared after the second split. But this is not enough to prove that the splitting scheme is wrong. That could also be the cost actively paid by the team after weighing options to make all FPGAs work simultaneously.

Jiang Lin needs to see how long each task block actually ran, how much duplicate computation was generated, and how much cross-node communication was produced, in order to judge whether the price they paid exceeded the benefits brought by load balancing.

He wrote down the second set of opinions.

[Preliminary screening: The anomalies are concentrated in the second-stage partitioning phase of "packing by category first, then splitting to the chips".]

[Current materials are insufficient to judge whether there are omissions in the splitting method or if it is the price actively paid by the team to evenly distribute computing power.]

[Supplementary requirements: Actual execution time of each task block, scale of similar tasks, duplicate calculation rate after splitting, cross-chip verification rate, and summary communication volume.]

[Status: Project not established for now.]

The third item, surface code syndrome events.

Jiang Lin lingered the longest on this item.

This time, you can imagine a quantum chip as a warehouse that people are not allowed to enter directly.

Extremely fragile goods are stored in the warehouse. A slight external vibration, temperature change, or control deviation can cause a problem with a certain piece of goods.

Yet the staff cannot open the door to inspect piece by piece.

Once quantum information is directly measured, it may be destroyed. Many times, just as people confirm what state it is in, they simultaneously destroy the thing they originally wanted to protect.

The surface code method is to deploy a large number of alarms around the warehouse.

The alarms will not tell the staff that the second piece of goods in the third row is damaged, but will only give the inspection results of a small area.

The complete readings given by all alarms in each round can be roughly understood as a syndrome. The locations that change between two adjacent rounds will be recorded as detection events.

A single detection event cannot directly locate the fault.

Researchers have to piece together the locations, chronological order, and durations of many detection events in order to deduce what most likely happened.

If all faults were independent of each other, things would be much simpler.

One ring in the east and one ring in the west can be treated separately as two unrelated minor faults.

However, errors on real quantum chips often appear in pairs or groups.

A crosstalk can cause adjacent areas to alarm simultaneously. The drift of a shared readout line can cause multiple channels along the line to produce anomalies together. The leakage of a certain qubit can also continuously affect several subsequent rounds of inspection.

If the decoder still splits these alarms into irrelevant minor events, it might find the wrong true fault path.

What is even more troublesome is that multiple alarms ringing simultaneously does not necessarily mean that a chain fault has truly occurred in the warehouse.

They might just be connected to the same line, might have suffered the same clock skew, or might have just happened to coincide with device recalibration. Truly correlated faults and illusions created by the recording system will appear in the data with nearly identical appearances.

The USTC team submitted two million alarm records.

Timestamps, detection rounds, anonymous channel numbers, and error labels are neatly arranged row by row.

But what Jiang Lin saw was an alarm log without a warehouse floor plan, without line groupings, and without a calibration schedule.

Two alarms always ringing together might be because they are closely adjacent to the same fault, or perhaps just because they share a single line.

A certain alarm being anomalous at fixed time intervals could be because the quantum error has periodicity, or perhaps just because the system is executing routine calibration at that moment.

Lacking this background information, two million records will not automatically turn into two million pieces of valid evidence. The more complex the analytical model, the more likely it is to interpret a non-existent physical fault from the traces left by the data acquisition system.

Jiang Lin did not download the data entities, but only wrote beneath the summary:

[Preliminary screening: The current data cannot yet determine whether the patchy alarms come from real faults or illusions created by the recording system.]

[Supplementary requirements: Anonymized channel adjacency relations, inspection types and measurement roles, experimental round boundaries, readout line groupings, calibration periods, frame loss and reset markers.]

[Restrictions: Supplementary information must still be desensitized; absolute device positions, process parameters, and control waveforms are not required.]

[Status: Data entities not accepted for now.]

The three entry reviews concluded, and the time had already reached 9:57.

Jiang Lin has currently only completed the processing level judgments.

Any of these items truly unfolding will require going through material re-examination, independent reproduction, counterexample testing, and external verification. The several opinions written tonight are still very far from deliverable conclusions.

He created a new template.

[External_Request_Minimum_Form.md]

[Background description, limited to two hundred characters:]

[Work already completed:]

[Current failure samples or anomaly objects:]

[Reproducible materials and data formats:]

[Single question requested for Jiang Lin to judge:]

[Physics, confidentiality, and ownership boundaries:]

[Matters explicitly not requesting assistance:]

[Expected delivery interface:]

Jiang Lin sent the template back to the Research Support Unit, accompanied by processing instructions.

[Instruction 1: BB(5) enters reproduction preparation. Request the version numbers, operating environment, minimum test set, and all unified failure samples of the three types of deciders. Do not engage in joint development with the other party at the current stage.]

[Instruction 2: D(9) project not established for now. Request task block statistics according to the supplementary list; do not discuss refactoring schemes before materials are completed.]

[Instruction 3: Quantum error correction data not accepted for now. Return only the metadata supplementation list, still to be relayed by Professor Lu Zhixing.]

[Instruction 4: Subsequent external technical requests uniformly use Minimum Form. Those with incomplete formats are to be returned at the Support Unit.]

After sending the email, Jiang Lin established three isolated folders under the archive directory.

[BB5_NonHalting_Certificates / REPRODUCTION_PENDING]

[Dhain_Verification / SUPPLEMENT_REQUIRED]

[Syndrome_Correlation_Ledger / METADATA_REQUIRED]

Establishing the folders only indicates that the materials have been archived, not that all three problems simultaneously become his tasks.

Currently, the only one truly entering the processing window is the first item.

Jiang Lin clicked open the BB(5) directory and established three empty projects inside.

[Cyclers_Reproduction]

[Translated_Cyclers_Reproduction]

[Backward_Reasoning_Reproduction]

Subsequently, he glanced at the absolute sealing zone outside the peer-level directory.

The FDSO, TM-7, and StarTide preparatory engineering indexes were all there, each with independent numbering, verification records, and access restrictions.

They had no real project numbers, nor did they belong to the external problem queue.

Jiang Lin re-locked the absolute sealing zone and pulled his attention back to the center of the screen.

The first step does only one thing.

Confirm what the certificates submitted by the three types of deciders are each actually proving.