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.
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.
+Implementation status in the following additions is recorded at 2026-10-16 13:40 UTC, 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 reuses that goal projection in the planning horizon, showing the original Objective, provenance, separate acceptance and readback before work selection. A source correction appears on the next read. It creates no second persisted goal and does not rely on repeating a slogan every Turn.
+Both remain unmerged proposals at this cutoff. Ordinary installed-package and real-CLI checks verify source delivery, correction readback and unchanged selection authority; 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.
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. The intervention reduced mean fills 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, while drawdown worsened in ten of 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 local solver used about 25.25 CPU seconds; 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.
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.
+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.
2026-10-16 update: checkpoint continuation and trajectory diagnosis under frozen sources, not a new complete benchmark ranking. Historical versions and figures remain distinct. All four current research workers were deliberately stopped; scores, trajectories and archives were retained.
+A major recent engineering result is making mid/late research branchable and inspectable. Native execution history, LoopX research state, workspace code, incumbent, public feedback, submission opportunities, cooldown and cumulative budget must be saved at one safe boundary. Code alone loses the reasons for reaching this state; conversation alone loses executable state; successful snapshot creation does not establish restoration and useful continuation.
+Joint archive and recapture, complete preparation deadlines and fresh continuation before replanning progressively connect saving, restoration, evidence adoption and further work. Qualification includes isolated Linux, actual native HTTP/SQLite/Docker, ordinary installed packages and mature research points. Scripted models establish the mechanical chain, not model judgment. Concurrent receiver publication, native execution receipts and confirmed cleanup must also qualify together; one independently passing PR does not certify the final integration.
+One study-design correction was crucial: dump before replan, then let the fresh conversation make the new decision. Reviewing in the old conversation and asking a fresh one to execute that completed decision tests execution context, not fresh replanning; that option has been removed. Both conversations can access the same checkpoint research and original goal. The difference is retained native conversation history. Both arms enable stride; this is not a stride-on/off contrast.
+The latest Graph continuation pair used frozen LoopX source 7b8b08f8. Native scoring also penalizes solution time. Higher classification/calibration quality need not improve the total. The three bound complete evaluations pass all four checks across ten splits, with no resource penalty:
| 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 |
Quality gains of about 0.001524 and 0.000998 were outweighed by extra time penalties of about 0.004269 and 0.009406. Both arms retained the original best. This locates the next question in the full quality–time tradeoff rather than a higher classification submetric. Single-run timing still depends on conditions; it does not show that fresh context is inherently slower.
+A full-covariance candidate raised mean Sharpe across five public training proxies from about 1.394 to 1.504, yet native Sharpe fell from 1.3731 to 1.3272. Its score was 16.94 against inherited best 20.53. Drawdown and turnover components already earned 15/15 and 5/5 in this sample. Tightening those constraints does not automatically create alpha or risk-adjusted return. A negative result should separate failed transfer, inactive intervention and insufficient route potential before selecting the next experiment.
+At this stop, Portfolio’s retained-history arm had best legal score 20.54 and fresh had 20.53: a 0.01 increase above the inherited best for the former. Neither Graph arm exceeded its inherited best. Real research, valid native evaluation and incumbent preservation occurred; the small difference establishes no general superiority for fresh, retained or stride. Older fault-affected cohorts, this latest continuation pair and future repaired-source studies remain separate.
+ +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 |
| 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 |
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.
+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 next design 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 + stride. Qualify one task’s normal entry and first useful research before admitting the others. This is a proposed study; 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 | Portfolio and Graph first: same goal, evidence and version, stride on in both, fresh versus retained; preserve Ad points before expanding | Equal four-hour marginal windows, reporting common 1/2/4-hour outcomes; preserve consumed budget, opportunities and cooldown |
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.