--- title: "Same Shards, Fresh Window · example of Momentum · Agentic Atlas" description: "Same Shards, Fresh Window: why a fresh prompt beats restating requirements inside a conversation gone wrong." canonical: "https://agentic-atlas.dev/nodes/same-shards-fresh-window" last-updated: "2026-09-23" --- 1. [Agentic Atlas](https://agentic-atlas.dev/) 2. [Patterns in the Agentic Atlas](https://agentic-atlas.dev/atlas) 3. [Context Flow](https://agentic-atlas.dev/nodes/context-flow) 4. [Momentum](https://agentic-atlas.dev/nodes/momentum) 5. Same Shards, Fresh Window 1. example of [Momentum](https://agentic-atlas.dev/nodes/momentum) # Same Shards, Fresh Window **Can a fresh context window help an agent use the same information?** On the two models tested, a fresh window beat restating in place, since restating leaves the earlier replies resident, at the cost of condensing first. Hook why a fresh prompt beats restating requirements inside a conversation gone wrong Laws & fences - Restating down-weights the model's old attempts, but only removing them from the window takes them out entirely. - The 95.1% figure is a never-lost prompt's score relative to the full prompt, not what a lost session recovers. - Evidence about restarts should hold the information fixed and vary only whether the old replies remain. - The study supports condensing a session before retrying, not a guaranteed gain from every restart. When to reach - Reach for this when a stuck session tempts you to paste the full requirements and simply continue. - Reach for this when a quoted number claims restarts recover most of a session's performance. Provenance [same-shards-fresh-window/implementation](https://agentic-atlas.dev/nodes/same-shards-fresh-window#implementation) · v1.0.11 Addresses atlas_cards same-shards-fresh-window 2. [Momentum](https://agentic-atlas.dev/nodes/momentum) 3. field notes - “transform a payload to strengthen task-relevant signal in a smaller or more useful representation” The card, in place · its connections drawn edges from atlas_links same-shards-fresh-window On this plate [context](https://agentic-atlas.dev/nodes/same-shards-fresh-window#context) [problem-signal](https://agentic-atlas.dev/nodes/same-shards-fresh-window#problem-signal) [choice](https://agentic-atlas.dev/nodes/same-shards-fresh-window#choice) [before](https://agentic-atlas.dev/nodes/same-shards-fresh-window#before) [implementation](https://agentic-atlas.dev/nodes/same-shards-fresh-window#implementation) [result](https://agentic-atlas.dev/nodes/same-shards-fresh-window#result) [verification](https://agentic-atlas.dev/nodes/same-shards-fresh-window#verification) [lessons](https://agentic-atlas.dev/nodes/same-shards-fresh-window#lessons) [relationships](https://agentic-atlas.dev/nodes/same-shards-fresh-window#relationships) Every section is addressable on its own. Read only the ground you need. ## Context [Permalink to Context section](https://agentic-atlas.dev/nodes/same-shards-fresh-window#context) A long agent session that has gone wrong offers its operator two exits, and they look interchangeable: restate the full requirements into the transcript and continue in place, or consolidate what the session was told and [dispatch](https://agentic-atlas.dev/glossary/dispatch) fresh. [Momentum](https://agentic-atlas.dev/nodes/momentum) claims they are not interchangeable — the window steers, not just informs, and in-window correction re-weights a resident trajectory but cannot remove it. That claim is testable, and one study ran the test at scale. Laban et al. 2025, *LLMs Get Lost in Multi-Turn Conversation* (arXiv:2505.06120), took fully specified single-prompt tasks and **sharded** them: each instruction was split into fragments, each fragment carrying one piece of the specification, and the fragments were revealed one per turn across a simulated conversation. Fifteen models, six generation tasks. Four conditions matter here: - **Full** — the original instruction, complete, in one prompt. The reference bar: nothing withheld, nothing resident. - **Sharded** — the same content revealed shard by shard. The model replies every turn, so by the final turn the window holds the whole specification *plus* every intermediate reply. - **Concat** — the same shards concatenated into one fresh prompt. The same information as the sharded window's final turn; no conversation ever happened. - **Recap** — a sharded conversation plus one final turn restating every shard (run on two of the models). The same information again — delivered at the tail of the trajectory instead of in a fresh window. [Momentum](https://agentic-atlas.dev/glossary/momentum-context-momentum)'s claims predict an ordering across these four conditions; the study delivers it. ## Problem signal [Permalink to Problem signal section](https://agentic-atlas.dev/nodes/same-shards-fresh-window#problem-signal) The sharded conversations lost **39% on average** against Full, across the fifteen models and six tasks. Nothing was missing — by the final turn, every shard is in the window. What the sharded window has that Full's does not is a **trajectory**: the model's own replies, produced while the specification was still incomplete, resident and re-read on every subsequent pass. The paper's diagnosis matches momentum's: models attempt solutions early on assumed details, then over-rely on their own prior attempts — and once lost, they do not recover in place. The trap for a reader arrives with the second number. Quoted alone, "the same shards scored 95.1%" sounds like a recovery rate — as if a lost conversation was rescued to near-parity. It was not, and the two-panel reading below exists to make that misreading impossible. ## Choice [Permalink to Choice section](https://agentic-atlas.dev/nodes/same-shards-fresh-window#choice) Read the four conditions as **two panels**, each with its own declared baseline: - **Panel one — the force *(Full → Sharded)*.** Hold the information fixed; change only its delivery, from one prompt to a trajectory. The 39% drop is the price of a resident trajectory, not of missing content — Concat proves the shards carry the specification intact. - **Panel two — two exits from the lost state *(Recap vs. Concat)*.** Both deliver every shard. Recap delivers them at the tail of the lost conversation; Concat delivers them in a window where the conversation never happened. Residency of the trajectory is the only difference — the [Momentum](https://agentic-atlas.dev/nodes/momentum) run as an experiment. The rejected alternative is the single-panel reading, Sharded → Concat as "before and after." It compares a lost conversation to a control that was never lost, invites exactly the recovery-rate misreading, and hides the one arm — Recap — that isolates what restating in place actually buys. ## Before [Permalink to Before section](https://agentic-atlas.dev/nodes/same-shards-fresh-window#before) The baseline is the sharded trajectory at its final turn: ``` Full [complete instruction] ─▶ reply the 100% bar Sharded [s1]▶reply▶[s2]▶reply▶ … ▶[s_n]▶reply 39% below Full on average window at the end: every shard + every intermediate reply ``` The specification is complete; the score is not. The gap is carried by the resident replies — momentum's claim that mid-window course corrections *look* complete while the old heading is still in every forward pass, measured at scale. ## Implementation [Permalink to Implementation section](https://agentic-atlas.dev/nodes/same-shards-fresh-window#implementation) ### Shift 1 — restate everything in place *(Recap)* Keep the lost conversation resident and append one turn that repeats every shard — the study's version of the stuck session's reflexive move: paste the full requirements at the tail of the transcript and continue in place. Information completeness is restored at the recency-privileged tail; the trajectory stays. On the two models tested, Recap improved on Sharded — and stayed below both Full and Concat. That residual gap is the part of the loss that restatement cannot touch: **decay mutes; only [eviction](https://agentic-atlas.dev/glossary/eviction) removes.** The old attempts are re-weighted downward by the recap, not subtracted from the calculation. ### Shift 2 — same shards, fresh window *(Concat)* Discard the conversation entirely and hand the same shards to a window that never carried it — the consolidate-and-[redispatch](https://agentic-atlas.dev/glossary/re-dispatch) move, run under lab conditions where consolidation is free because the shards are already enumerated. One decision changes against Recap: whether the trajectory is resident. Concat averaged **95.1% of the fully specified control** — near the bar Full sets, with the trajectory's price almost entirely unpaid. A fresh dispatch is not merely full [relevancy](https://agentic-atlas.dev/glossary/relevancy); it is **zero momentum**. ## Result [Permalink to Result section](https://agentic-atlas.dev/nodes/same-shards-fresh-window#result) The four conditions order exactly as momentum predicts: ``` Sharded < Recap < Concat ≈ Full lost lost, then never lost never sharded told everything (95.1% of Full) (the 100% bar) ``` **What 95.1% is:** the Concat control's average against Full — the score of a single-prompt condition whose window never held a trajectory. **What it is not:** a recovery rate. No arm in the study got lost and then scored 95.1%; the arm that got lost and received the complete specification anyway is Recap, and it sits below. The number quantifies what a lost session's content is *worth once the trajectory is gone* — which is precisely the emergency reset's warrant: consolidate the shards, dispatch fresh. ## Verification [Permalink to Verification section](https://agentic-atlas.dev/nodes/same-shards-fresh-window#verification) Every figure traces to Laban et al. 2025 (arXiv:2505.06120) through [Momentum](https://agentic-atlas.dev/nodes/momentum)'s grounding pass (URLs fetched 2026-07-04; the Concat-control reading corrected 2026-08-07). Check the example against the source by confirming three placements: 39% is Sharded against Full; 95.1% is Concat against Full; Recap sits between Sharded and Concat on the two models tested. Bounds on what the panels prove: - The result supports *consolidate before retrying*; it does not measure a universal restart dividend. The shards here are compact and enumerable — a real session's consolidation step is itself fallible, and its cost is the *transform a payload to strengthen task-relevant signal in a smaller or more useful representation* [seam](https://agentic-atlas.dev/glossary/seam)'s, not this study's. - The Recap arm covers two models, so its placement is ordinal evidence, not a population estimate. - The study never uses the word *momentum*. The example asserts what the conditions isolate — a resident trajectory steering continuation — not any particular circuit behind it. ## Lessons [Permalink to Lessons section](https://agentic-atlas.dev/nodes/same-shards-fresh-window#lessons) The load-bearing comparison is the one that holds information constant and varies only residency: Recap and Concat hear the same shards, and the window that never got lost wins. Every other contrast in the study moves two things at once; this one isolates the trajectory, and it is the comparison to demand of any future evidence about resets. The shards are furniture. What travels is the fork and its prices: restating into the transcript pays the trajectory's price and keeps it resident; consolidating and dispatching fresh pays a distillation cost and zeroes the trajectory. And when a quoted number promises that restarts recover 95% of anything — ask whether the window behind the number was ever lost at all. The relationships ledger Evidence-bearing references ## Relationships Every connection keeps the section where it was found. The map above orients; this ledger carries the evidence. ### Outbound references 4 1. in-slice · occurrence 1 [Momentum](https://agentic-atlas.dev/nodes/momentum) how context already in the window steers the model along its current path Evidence: [Context](https://agentic-atlas.dev/nodes/same-shards-fresh-window#context) · occurrence 1 2. in-slice · occurrence 1 [Momentum](https://agentic-atlas.dev/nodes/momentum) how context already in the window steers the model along its current path Evidence: [Choice](https://agentic-atlas.dev/nodes/same-shards-fresh-window#choice) · occurrence 1 3. in-slice · occurrence 1 [Momentum](https://agentic-atlas.dev/nodes/momentum) how context already in the window steers the model along its current path Evidence: [Verification](https://agentic-atlas.dev/nodes/same-shards-fresh-window#verification) · occurrence 1 4. undisclosed · occurrence 2 Undisclosed relationship transform a payload to strengthen task-relevant signal in a smaller or more useful representation Evidence: [Verification](https://agentic-atlas.dev/nodes/same-shards-fresh-window#verification) · occurrence 2 ### Inbound references 1 1. in-slice · occurrence 1 [Momentum](https://agentic-atlas.dev/nodes/momentum#evidence-the-grounding-pass) why a fresh prompt beats restating requirements inside a conversation gone wrong Evidence: [Evidence](https://agentic-atlas.dev/nodes/momentum#evidence-the-grounding-pass) · occurrence 1 [↑ back to the top](https://agentic-atlas.dev/nodes/same-shards-fresh-window#content) [← the survey](https://agentic-atlas.dev/atlas) Node same-shards-fresh-window · corpus 78c0e17 · Catalog revision e0cb75881244b1a82193ca738b82a0d508dd62e822524ae82873ec79d51dbb61