Research question and comparison methods updated October 10, 2026. This revision adds no experimental scores; all numerical results below retain their stated historical versions and observation windows.
We want to test whether, with the same model and comparable budgets, a general harness helps an agent accumulate and use experience well enough to keep making verifiable progress later in execution. A failed route should leave more than “failed before”: its conditions, what it ruled out and what new evidence would justify retrying. Recording that finding, changing a subsequent choice and improving the outcome require separate checks.
A main session that continues to interpret the goal and evidence is one way to study this question; cross-session recovery tests reliable state transfer. Session continuity alone does not establish better use of experience. Fresh-context execution and workflows can also preserve experience through files and external state. Comparisons should record the actual session strategy, available information and handoff cost, letting results determine applicability. EdgeBench Section 5.2 and Appendix G provide relevant comparisons of sustained execution, independent retries and harnesses; they do not isolate LoopX’s incremental benefit.
+The 2026-10-16 continuation is integrated below: goals and recovery, evidence follow-through, whole-episode cost, planned/stride, pre-decision checkpoints and real contrasts, and the next study design. The opening and earlier numbers retain their original cutoff.
-Portfolio requires a strategy developed on public training data and consistent trading, risk and accounting outputs. Lean requires proofs that preserve fixed declaration constraints and pass compilation and axiom-dependency checks. Negative evidence in Portfolio may mean a candidate is not worth pursuing under specified conditions. In Lean, it may mean a statement lacks an assumption and the proposed proof is impossible. Both demand sustained work, but their evidence and verifiable outcomes differ.
+The research has two stages. Early Portfolio risk calibration and Lean proofs tested trustworthy feedback, sustained work and evidence adoption. Recent work uses mature Portfolio and Graph states to study later choices and prepares restorable Ad Placement points. The table preserves the early five execution modes; it is not the current rerun matrix. The latest same-point comparison enables stride in both fresh and retained arms. The next design starts one LoopX donor per task.
| Execution mode | Mechanism retained | Question |
|---|---|---|
| Official worker | Official stop hook and continuation | How far does the official execution path progress? |
| Single Codex call | One call, ending naturally | How much happens without an external continuation loop? |
| LoopX Explore | Resume plus optional evidence capture and exploration planning | Does structured evidence enter the next decision? |
The historical Portfolio design covered five workers × native/blind feedback, ten arms. Later work added selective ablations. Lean primarily uses blind runs. Native workers may obtain judge feedback. Blind workers retain public tasks, training data, local tests and validators, but receive neither judge scores nor hidden diagnostics. Background evaluation is an operator observation and is not returned to the blind solver. Blind means no judge feedback, not no feedback.
-The reported model is gpt-6.1-sol / xhigh. Task defaults in the interim study were a 12-hour ceiling with five-minute sampling for Portfolio, and 24 hours with 30-minute sampling for Lean. Portfolio had no fixed aggregate evaluation concurrency cap; Lean had an independent cap of two. Earlier runs used 18 hours, and selected Lean runs were extended to 36 hours. Those original settings remain recorded rather than being rewritten to match new settings or presented as live defaults for later versions.
We also explored seeded phase Todos versus LoopX-generated plans, review cadence, context delivery and Explore changes. These attempts accompanied prompt, runner and contract repairs. Planned entry looked more promising and became the benchmark default. That is an engineering choice based on current evidence, not a universal advantage established by single cross-version runs.
-Lean provides direct common-window evidence. Under blind conditions, valid samples from early LoopX v14 and current #5844 runs are shown below. More verifiable proof progress in the same time is a real observation, even before its component causes are separated.
-| Lean raw score | Early Resume v14 | Current Resume | Early Explore v14 | Current Explore |
|---|---|---|---|---|
| 180 minutes | 0.1103 | 0.3478 | 0.1134 | 0.4019 |
| 450 minutes | 0.1882 | 0.5243 | 0.1996 | 0.5605 |
In these single-run windows, the current LoopX paths reached roughly 2.8–3.5 times the earlier raw scores. This ratio describes observations. Prompts, task entry, runners and control mechanisms changed together; it is neither the causal gain of one feature nor a result to extrapolate directly to other tasks.
-Portfolio improved differently. Early persistent zeros included feedback-wrapper and accounting-contract defects. After repairs, blind arms produced valid ledgers, conducted local research and obtained usable evaluations. Older Resume reached a best sampled 17.32; later planned v17 reached 25.39, but its last sample was only 12.38, versus 15.57 for older Resume. Budgets and scoring revisions also differed. At roughly 7h45, current #5844 Resume/Explore had best sampled scores of 15.01/12.21 and latest samples of 8.15/9.67. Evaluability, continuation and some best observed outcomes have moved beyond early obstacles; terminal quality and sustained retention of good solutions still need work.
+The model remains gpt-6.1-sol / xhigh. Historical Portfolio/Lean figures retain their budgets, sampling, scorers and dates; earlier 18-hour or extended 36-hour runs are not rewritten. Recent Graph/Portfolio continuation preserves consumed time, opportunities and cooldown within a cumulative 36-hour ceiling. That is not a fresh 36-hour allowance for every new donor. The next design uses task limits of 12/18/18 hours, detailed below.
Earlier work compared seeded phase Todos with goal-derived planned execution, review cadence, context and Explore. Later research selects planned + Explore and explicitly reviews every six settled effective work Turns. This is the current study setting, not a claim that old three-Turn runs used it or that explicit product units changed. Prompts, runners and contracts co-evolved, so early planned results do not establish a general advantage on their own.
+Lean provides direct common-window evidence. Under blind conditions, valid samples from early LoopX v14 and that-stage #5844 runs are shown below. More verifiable proof progress in the same time is a real observation, even before its component causes are separated.
+| Lean raw score | Early Resume v14 | #5844 Resume | Early Explore v14 | #5844 Explore |
|---|---|---|---|---|
| 180 minutes | 0.1103 | 0.3478 | 0.1134 | 0.4019 |
| 450 minutes | 0.1882 | 0.5243 | 0.1996 | 0.5605 |
In these single-run windows, the then-current LoopX paths reached roughly 2.8–3.5 times the earlier raw scores. This ratio describes observations. Prompts, task entry, runners and control mechanisms changed together; it is neither the causal gain of one feature nor a result to extrapolate directly to other tasks.
+Portfolio improved differently. Early persistent zeros included feedback-wrapper and accounting-contract defects. After repairs, blind arms produced valid ledgers, conducted local research and obtained usable evaluations. Older Resume reached a best sampled 17.32; later planned v17 reached 25.39, but its last sample was only 12.38, versus 15.57 for older Resume. Budgets and scoring revisions also differed. At roughly 7h45, that-stage #5844 Resume/Explore had best sampled scores of 15.01/12.21 and latest samples of 8.15/9.67. Evaluability, continuation and some best observed outcomes have moved beyond early obstacles; terminal quality and sustained retention of good solutions still need work.
This supports a more effective overall system while leaving two questions open: where did the gains come from, and how do we prevent regression with a new version or more runtime? Improving benchmark execution—shared iteration guidance, correct recovery, task-specific sampling and budgets—is itself part of long-horizon systems engineering. Its contribution should be recorded rather than attributed entirely to the graph or replanning.
One concrete failure occurred when seeded-blind wrote no_followup after roughly 34 minutes. The outer process resumed about 100 times without restoring useful work. A live scheduler is not a runnable semantic path. In another failure pattern, the system already knew replanning was needed but first offered a predictably invalid choice, forcing a failed call before revealing the next step.
No existing Todo does not imply that the original goal is complete. After finishing a work item, the agent must compare the result with the original acceptance: is it satisfied, does a reasonable next step remain within authorization, is there an external dependency, or is there truly no useful continuation? Replanning may produce work or justify staying on the same route. It should not invent tasks merely to keep a loop moving.
The previous product default checked periodically after five completed Todos. A proof task can stay unfinished for hours, so Todo completion may lag actual investment. #5879 changed the implicit unit to settled effective work Turns: five for the product and three for benchmarks at that revision, while explicit legacy Todo settings retained their unit. The change has merged. Frozen experiments retain their original settings, and a source merge does not automatically upgrade installed instances.
+A long task can remain unfinished for hours, so review should track actual settled work. The effective-Turn repair replaced implicit Todo-completion counting while preserving explicit legacy units. Current continuation and the proposed study use six effective work Turns; early three-Turn and then-product five-count settings describe their own revisions. A due review requires reconsideration, not a route change or new budget.
A Turn here is neither a model message nor a tool call. Cadence comes from uniquely settled work: uncommitted refreshes, retries and duplicate records must not inflate the count. Reaching the interval asks for a review, not a mandatory route change, and grants no budget or authority. Real entrypoints have been tested to trigger while a Todo is still open and to resume work after a justified continuation. Net benefit depends on what the model actually decides.
The counter must also distinguish work from observation. A validated negative result can be effective work; monitor-only polls, duplicate submissions and unsettled attempts do not count. This lets a long-open Todo receive periodic review without letting background sampling trigger pointless replans. A review should connect the original goal, candidate evidence and the next action. Choosing to continue can be correct; acknowledging that a check occurred does not establish better research direction.
Review cost also needs a breakdown by trigger. Periodic checks, a remaining goal gap after a work item ends, and new evidence invalidating a key assumption may need different depths of review. Changing the interval affects only one category. If every small completion rebuilds the global plan, a longer interval may not solve the problem. Conversely, a replan phase can contain useful probes and validation, so its entire duration is not administrative waste. Traces should distinguish useful route revision, repeated evidence gathering and mechanical bookkeeping before testing the boundary between local continuation and heavier global review.
Session recovery preserves execution context. Settlement connects the current Turn, selected work, validation result, durable writeback and budget use. A retry after a lost response needs to recognize existing effects rather than execute or count them twice. Restoring a session without settlement can revisit the same obstacle; correct settlement with the wrong session can lose the solving context.
-Repairs followed existing owners: exact session recovery, receipt-scoped next actions, and ineffective-continuation diagnosis. These have concrete failure-repair evidence. They establish that work can continue correctly, not that the next work is well chosen.
+Earlier repairs addressed exact session recovery, next-action/receipt binding and invalid continuation diagnosis. Recent work adds checkpoint-selection and lifecycle recovery. Resume from the original Turn and committed effects: a lost reply must not duplicate execution or spend, and independent new work requires a fresh legal-path judgment. Correct continuation and useful work selection remain separate acceptance questions.
+The following additions are updated through 2026-10-16, distinguishing merged repairs from proposals under review without rewriting earlier windows.
+Recent inspection moved beyond whether an agent read a Goal to which goal actually arrived. The original task remained in prompts and files, and agents did read it; some newly created structured benchmark Goals nevertheless carried only generic research reminders. Work could satisfy that reminder while drifting away from the task’s main metric. A separate projection defect borrowed public-display filtering for scoped input: paths or secret-like strings could drop a whole goal paragraph, and research text could be truncated.
+#6649 preserves the original task objective and priorities in new Goals and complete source text in authorized same-Goal replan input; public export boundaries remain separate. #6654 makes that shared goal projection a default-off, explicit opt-in in LoopX and its benchmark runner. When enabled, the planning horizon shows the original Objective, provenance, separate acceptance and readback before work selection. A source correction appears on the next read. Disabling it preserves the existing horizon behavior; the Goal input already supplied to replan is independent of this switch. It creates no second persisted goal and does not rely on repeating a slogan every Turn.
+The original-goal repair #6649 is merged; horizon proposal #6654 still awaits independent review. The switch revision has configuration-editor and CLI readback, feature-off parity and benchmark effective-setting checks; the shared projection also has installed-package and real-CLI source-delivery validation. These validate candidate source, not its adoption by the research integration; old checkpoints are not rewritten. This repairs visibility, not goal prioritization by the model, and cannot explain every earlier low score.
+In the older Graph stride-on/off comparison, outer wakes failed 20/99 and 14/104 times; successful wakes also recorded 35 and 45 internal stage errors. Failures clustered around review, checkpoint selection and unsettled-state recovery. The outer runner could wake again, while committed candidates, native results and some checkpoints survived. Work therefore continued, but could repeatedly recover the same decision, delay useful action and consume research or capture time.
+Recorded failed-stage time was about 49 and 35 minutes, not total loss; 20/99 does not mean 20% of useful work was wasted. Unequal fault burdens confound an overall comparison. The selection-hop repair and lifecycle recovery repair address distinct breaks: preserve the original Turn, recognize committed effects and restore a legal route when independent work arrives. They add no submissions or cooldown resets and do not retroactively convert failures into passes. New studies need actual adoption of the repaired source.
One supplementary Portfolio sample records a risk finding, then implements and tests tail protection. On this period’s actual inputs, however, constrained and unconstrained allocations remain equal. A finding reaching code and passing stress checks does not establish an intervention in this task. A Lean chain proceeds from periodic-review candidate declarations to installing proofs, a full build and later batches. That is closer to adoption evidence than a node count, but still does not exclude the same conversation and local files supplying equivalent information.
A boundary failure is informative too. A supplementary Lean run rejects a ninth question association for one Todo because the active association set is limited to eight. It preserves existing associations, records the extra conclusion as standalone Goal-scoped evidence and resumes ordinary work. This limits one task’s association set, not the whole graph to eight findings. Recovery is not free. Reducing capture friction should preserve relevant decision evidence rather than silently deleting old counterevidence to make a write succeed.
+Evidence existing in a checkpoint, a model having read it, and current replanning being able to use it structurally are different facts. We saw a new Goal’s Explore projection report an empty current graph while the archive retained historical research. That requires source/adoption repair, not concluding that experience is absent. Later continuation qualification checked research state, the original goal and actual Explore delivery; mature examples contain dozens of work Turns and over a hundred evidence events. This supports recoverability, not research quality by record count.
+In one historical Graph audit, five of ten subsequent reviews still cited an older decision. This can skip the experiment just completed and keep reviewing a stale route. Repair in the existing history-sequence and stride owners retrieves the immediate predecessor in all ten prefix replays and rejects five stale references. It establishes the right review target; the model must still judge execution, new information and progress toward the original goal.
+A discriminating record connects previous choice → actual candidate version → completed experiment → public feedback → next adoption or rejection. Explore’s incremental value belongs on that chain: which condition did it preserve, repetition did it avoid or choice did it change? A node, a switch label or a branching graph alone proves none of those.
Exact Todo output fell from 15,986 UTF-8 bytes and three body copies to 2,721 bytes and one copy. Ordinary quota grew from 23,088 to 27,137 bytes against the command-only version because it carried actual content. Pending sources shrank from preference + Goal + Todo to Goal. This does not establish lower total tokens or latency: if old reads were already batched, removing a CLI process may remove no model round trip.
In 12 frozen real wakes, all received the complete Todo: nine through inline or file-based guards, three through exact reads. Eighteen inspected bodies matched their startup source digests. That establishes delivery and source consistency, not comprehension, complete constraint adherence or total cost benefit. The earlier roughly 78% synthetic compression claim likewise applies only to its measured output.
-Progressive access does not mean offering a sentence and making the model guess. It separates information by decision relevance: complete constraints for current work arrive directly; complete intent, history and additional evidence retain reliable entrypoints. Failed sources, changed selection or changed revisions require recovery and current context, not treating stale content as a successful read.
+The next integration’s information organization follows one real decision: complete current-work constraints arrive directly; a same-source original goal anchors selection; additional history and evidence expand by relevant question and revision. Do not mechanically reread an already delivered source. Missing content, failed sources or changed versions require complete reads through the current contract. Authorized research input must not lose its objective or counterevidence to public-display filtering. Optimize the whole decision path rather than one JSON length.
Optional capabilities must not leak through generic guidance. Repeated semantic-preference reads induced by a generic skill while the capability is off are a boundary defect; #5874 returns that guidance to the participating hook. Smaller packets and fewer useless prompts help, but must preserve acceptance tails, applicability, continuation conditions and failure recovery. The optimization unit is the decision path, not one JSON length.
A direction worth testing is to put state and recovery complexity inside semantically complete tools with clear boundaries: obtain all constraints needed for current work, for example, or submit a result with its source and applicability. A tool may perform several internal checks while the model remains responsible for interpreting evidence and choosing a route. Anthropic’s tool-design guidance likewise organizes interfaces around real work rather than exposing every underlying API to the model.
Fewer commands alone are not the objective. Consolidation must preserve permissions, visible failures and retry semantics. Real execution must show whether an interface is easier to learn, produces fewer errors and turns saved interactions into useful work. #6038 improves action-specific CLI help and recovery from lease-argument errors. That repairs a usage path; it does not yet establish better task outcomes or lower whole-run cost.
+A recent Portfolio fixed 2×2 check covered 14 local cases and 56 result rows: five public cases and nine synthetic cases. The factors were original V7 versus adaptive covariance, and truncating total target shares versus truncating the requested share change. The following means use only the five public cases, hold V7 fixed and change only the execution rule to share-change truncation. Mean fills fell from 451.2 to 235.4, about 47.8%; costs fell only from 62,683.15 to 62,682.18, about 0.00154%. Mean Sharpe rose just 0.000133. Separately, drawdown worsened in ten of all fourteen cases. A large change in operation count produced a tiny change in the goal. Fewer fills became a candidate mechanism whose value needed comparison with actual headroom.
+The solver reported about 25.25 CPU seconds for local computation; the two complete wakes containing this research took about 29.66 minutes, ran 90 commands and emitted roughly 68,000 output tokens. A separate completed covariance episode took seven wakes, about 68.05 wall-clock minutes and 144,700 output tokens. These include reasoning, control, authoring, tools and waiting, not pure decoding time. Some diagnosis carries information and is not all management waste. Reporting only a small script’s seconds nevertheless misses the main cost relevant to continuation.
+Measure the full episode, then distinguish active work, recovery, waiting and native evaluation. Difference and deduplicate cumulative session usage rather than adding ancestor history repeatedly. Faster local computation, fewer commands or smaller packets count as efficiency gains when they preserve decision information and improve useful output or complete cost.
The improved #5844 Resume and Explore arms progressed quickly early on, then were overtaken by historical native Goal in the middle. We analyzed common 180–360 minute windows using solver-visible tool results and frozen source snapshots. Current arms share c0d519cf; historical Goal used c46f397c. Model and public task digest match, but dates, runners, prompts and budgets differ, with one run per cell. This is route diagnosis, not a clean causal framework comparison.
The improved #5844 Resume and Explore arms progressed quickly early on, then were overtaken by historical native Goal in the middle. We analyzed common 180–360 minute windows using solver-visible tool results and frozen source snapshots. Those arms share c0d519cf; historical Goal used c46f397c. Model and public task digest match, but dates, runners, prompts and budgets differ, with one run per cell. This is route diagnosis, not a clean causal framework comparison.
| Blind arm | Raw score at 180m | Raw score at 360m | Net sorry reduction | Files affected |
|---|---|---|---|---|
| Resume / planned / #5844 | 0.3478 | 0.4919 | 502 | 17 |
| Explore / planned / #5844 | 0.4019 | 0.5147 | 442 | 18 |
| Change | Demonstrated benefit | Technical substance | Not yet established | |
|---|---|---|---|---|
| Task/evaluation validity | Strong: reproduced false rejection/blind spots and separated prompt confounds | Event timing, path ledgers, independent reconstruction, versioned experiments | A fair ranking after repairs | |
| Continuation/settlement | Strong: concrete ineffective recovery and retry defects are testable | Separate goal and Todo completion; session, result and settlement identities | Long-run net benefit and failure rates across tasks | |
| Context delivery | Local: deduplication, full body delivery and off-state isolation | Canonical content once, progressive reads, shared ordinary/envelope owner | Net tokens, latency and constraint-retention improvement | |
| Work-turn review | Mechanism: triggers while a long Todo stays open; retries do not count | Settled-fact counting; review distinct from switching; explicit legacy units | Better selection rather than more ceremony | |
| Explore evidence cycle | Observed: natural capture, association, later delivery and visible actions; actual intervention is not guaranteed | Scoped findings, stable question identity, task association, optional hooks | Incremental value over files/conversation and net task-quality or cost gains |
| Boundary | Currently verified | Still unproven |
|---|---|---|
| Goal and feedback | Reproducible ledger misjudgments repaired; full Goal input and same-source horizon proposals have real CLI/installed readback | Actual adoption, task prioritization and net benefit; old checkpoints are not migrated |
| Research state and recovery | Joint capture, fresh before review, original identity/opportunities/cooldown/deadline, mature-point continuation and further work are inspectable | Sustained stability as points grow, complete historical cost and a complete user-facing restoration journey |
| Evidence and review | Conditional capture, later delivery, immediate-predecessor review and real behavior chains; negative results and stale citations preserved | Explore beyond session/file memory; correct references yielding better choices |
| Planned and stride | Merged initial-quality/later-gain distinction; questions about retained best, goal bottleneck and actual implementation | Initial quality, later gain and cross-task stability after new-source adoption |
| Effect and efficiency | Graph quality gains outweighed by time penalty; Portfolio transfer counterexamples and a small best gain; episode costs decomposed | Repeatable net benefit in clean matched studies, beyond local speed or repair counts |
These experiments provide positive grounds for pursuing this direction: more verifiable blind work, fewer broken continuation loops and increasingly observable connections between evidence and tasks. A control plane should make “work happened,” “the result is trustworthy” and “the next step has a reason” separately inspectable. A break anywhere can leave the system active but ineffective. Connecting them still requires a model capable of choosing research directions. Progress is moving from keeping an agent running longer toward turning more runtime into useful work. The next stage is to stabilize that system-level improvement and test whether it can raise the capability ceiling further.
+Early system progress supports continued research. Recent work moves the question from sustaining execution to using early experience for better later choices. The changes and cases below separately assess inputs and judgment mechanisms, actual adoption, native outcomes and full cost. These levels cannot substitute for one another.
+The recently merged #6621 restores an important distinction in shared planning: deliver a strong, coherent first artifact feasible within the budget. Do not manufacture a weak start and call repayment of that quality deficit long-horizon capability. Useful diagnostics remain legal, and preparation cannot postpone delivery indefinitely. Later improvement is anchored to the retained best artifact.
+Stride keeps its three axes: work scope, route distance and validation scope. Its questions more explicitly compare the remaining authoritative gap, limiting component and plausible headroom; check actual implementation of the proposed mechanism; and distinguish an unexecuted change, failed transfer and insufficient route potential. For a candidate’s own screening threshold, ask what goal it protects, how the incumbent behaves under the same control and whether the discrepancy warrants vetoing an informative native evaluation. These guide judgment rather than force a numerical algorithm switch every few Turns.
+The trajectories also contain more than small parameter changes: full covariance, class-prior and within-class ranking approaches differ in mechanism. The unresolved issue is whether they affect the factors limiting native reward and complete an informative test at proportionate cost. Better questions, correct evidence and larger experiments may help; actual adoption and later outcomes establish their value. Frozen old runs were not hot-updated, so a source merge cannot be credited with changing their choices.
-As recorded at 2026-10-08 09:33 UTC, the supplement consists of Explore × native/blind on one frozen mainline: two Portfolio and two Lean arms. Each task also has one official blind workflow repeat; existing Resume, Explore and initial baselines are reused. Portfolio uses 12 hours with five-minute sampling; Lean uses 24 hours with 30-minute sampling. The LoopX supplements use planned entry, review every three settled effective work Turns and keep TurnEnvelope off. Exact source, image and original starts remain distinct; later repairs do not hot-update old arms.
-| Closeout item | Question | Decisive evidence | ||
|---|---|---|---|---|
| Portfolio: terminal quality and retention | Do training selection, risk intervention and final quality agree? | Reproduce the best sample and final artifact separately; independently reconstruct the ledger and test original return/risk goals and actual intervention, retaining failures | ||
| Lean: valid progress and later selection | Does later work still validate original declarations, and does evidence alter candidate allocation? | Standard axiom and original-type checks, capture-aligned windows, review → implementation → validation chains and complete phase costs | ||
| Official repeats | How much do baseline routes and validation resources vary? | Keep historical arms and compare common windows and terminal outcomes; one repeat does not establish statistical stability | ||
| Isolated replay of existing frozen state | Could an unresolved mechanism ambiguity change the conclusion? | Reuse identical source and work only when needed; do not expand a confounded long-run matrix |
| Graph raw score | Quality term | Time penalty | Final score | Mean solution time |
|---|---|---|---|---|
| Inherited incumbent | 0.753764 | 0.018622 | 0.735142 | 6.68 s |
| New retained-history candidate | 0.755289 | 0.022892 | 0.732397 | 10.47 s |
| New fresh candidate | 0.754762 | 0.028028 | 0.726734 | 17.89 s |
At that cutoff, the Portfolio supplements had ended and received independent numerical and ledger reproduction; complete Lean and official-repeat closeout was still pending. This methods update does not present that historical status as live runtime state or add unverified terminal results. Final analysis must reconcile solver termination, all captured scoring, integrity, original task goals, whole-run cost and evidence adoption. A terminal score returning does not mean every earlier capture has finished scoring; a successful source build does not mean every original declaration meets acceptance.
-An integrated change tests whether the system is more useful; independent repeats test stability; ablations test what caused a difference. Each requires different evidence. More tasks or execution modes do not substitute for repeated runs under the same conditions, and one best result does not establish a stable gain. Further comparisons should fix model, task and evaluator revisions, prompts, feedback access, session strategy and budget. Integrated exploration can identify the main differences, followed by repeats of frozen configurations and a small number of discriminating ablations. Tasks used for development and repeated configuration selection should be distinguished from those reserved for confirmation.
-A strong baseline should retain normal continuation, file-based notes and legitimate feedback. Testing explicit evidence organization requires comparison with files or conversations that can retain equivalent information. A protocol that returns a candidate’s full permitted feedback only when it establishes a new best is a separate feedback condition: define how best is established, what may be returned and when notifications occur in advance, then compare harnesses within that condition. Richer feedback or a different notification frequency cannot be credited to structure alone; silence does not imply task regression.
-| Comparison dimension | Evidence to retain together | |
|---|---|---|
| Quality and retention | Best valid artifact and final usable delivery at the same budget, including degradation; a best sample does not replace terminal acceptance | |
| Time and cost | Time and expense to comparable quality; model, tool, control and recovery costs; evaluation cost and resource waiting reported separately | |
| Later gains and mechanism | Capture-aligned useful work later in execution; links between evidence delivery, reading, changed choices and actual actions | |
| Stability and reproducibility | Independent-repeat distributions, failures, invalid runs and manual intervention; whether another user can obtain artifacts and explanatory records on the specified version |
| Observation | Sample and verification scope | What it does not establish |
|---|---|---|
| Graph quality and time penalty | One incumbent and two new candidates, three complete public-feedback reports; ten splits each, quality minus time penalty equals final score, zero resource penalty | A single timing difference is not a stable causal context effect |
| Portfolio fixed 2×2 | Four variants across fourteen cases; reported means for the V7 execution-rule contrast use five public cases; drawdown counts use all fourteen | Local proxy gains do not establish native transfer |
| Old Graph failures and reviews | 203 wake receipts recounted by arm; eleven reviews minus the initial review without a predecessor leaves ten reference checks | A capture error does not prove a lost checkpoint; repair replay does not prove model adoption |
| Research-episode cost | Two complete Turns for the fixed study; seven complete Turns from covariance design through native feedback and rollback, inherited usage removed, active time and wall time distinguished | Script CPU time is not full research cost, and token counts are not prices |
Further results will continue in this bilingual article: gains scoped to tasks and conditions, regressions analyzed through choices and costs, and unresolved contributions kept uncertain. Enabling another researcher to inspect settings, understand negative results and reproduce key comparisons is what turns interim observations into cumulative research evidence.
+The original archives and recomputation materials were supplied to an independent reviewer but are not yet published with this article. Public readers can check its numerical relationships and sample scopes, but cannot independently reconstruct every run from the article alone. Source-repair PRs and a passing page build do not close that data-access gap. A later data release should preserve these source links and negative results, rather than only the best scores.
+Earlier Portfolio/Lean supplements retain their original records; they are no longer the active launch plan. The four recent Graph/Portfolio continuation workers have stopped, along with their controllers and owned services. Feedback and archives remain. Interrupted/unknown terminal receipts are not relabeled completed, and known individual native results do not automatically qualify the entire run. Historical capture integrity, cost and scientific closeout still require explicit reconciliation.
+New studies use the latest qualified integration, separately pinning LoopX, runner, native provider, scorer, task, image and ordinary package; old runs are not hot-updated. First qualify the complete original goal, research evidence, real restoration and useful work, then start model research. The design below replaces the old four-arm supplement plan, adds no cancelled official baseline and does not rescore and stitch old versions.
+Continuation can save the cost of accumulating early research, but repeatedly reusing fault-affected old points cannot establish the repaired product’s overall effect. The current proposal starts one fresh LoopX donor each for Portfolio, Graph and Ad Placement on the latest qualified integration, retaining gpt-6.1-sol / xhigh and using planned + Explore with stride off and ordinary review retained. TurnEnvelope stays off in donors and receivers; checkpoint capture is independent, and actual input is verified. Qualify one task’s normal entry and first useful research before admitting the others. This proposal is not yet frozen or launched; no new paid experiment was started for this update.
| Stage | Design and criterion | Budget and limits |
|---|---|---|
| Accumulate from scratch | One arm per task; pin task, model, scorer, installed source, feedback and prompt; inspect initial delivery and actual evidence adoption | Task limits of 12/18/18 hours, at most 48 worker-hours; no six-arm fresh matrix |
| Produce research points | Capture before natural replans; preselect safe boundaries near half and three-quarter time, targeting two qualified mid/late points per task | Six target points; no score-based selection, deletion of stalled/low-scoring observations or artificial replans to hit a clock |
| Compare the same point | Proposed first comparison: Portfolio and Graph, same goal, evidence and version, both replanning fresh, varying stride on/off only; compare fresh/retained separately | Equal four-hour marginal windows, reporting common 1/2/4-hour outcomes; preserve consumed budget, opportunities and cooldown; account separately for later context-arm costs |
A quality checkpoint is complete, attributable, captured before the current decision and actually restorable in an ordinary package. Sparse research is itself an observation; quality must not become post-hoc winner selection. Native submission has a one-hour cooldown from acceptance, is initially eligible and receives no extra tail grading. Review cadence is six settled effective work Turns throughout. A defect in goals, evidence, restoration or feedback stops the affected pair; preserve its reason and failure cost, then record a separate repaired attempt.
+Primary outcomes are legal native gain above inherited best, time to first gain and complete marginal cost. Process evidence asks how a negative result changes the next real experiment. Three fresh LoopX trajectories versus historical official runs remain descriptive. General LoopX-versus-native effects require separately admitted native repeats on the same version and budget; historical baselines cannot be stitched into a new causal control.