Kapan sampeyan kudu nyimpulake riwayat obrolan tinimbang ngirim maneh kabeh?

Question

Accepted Answer

Nyimpulake obrolan nalika riwayat dadi dawa lan rincian sing luwih lawas ora perlu maneh kanggo tugas saiki. Tinimbang ngirim maneh transkrip lengkap saben giliran, sampeyan ngganti bagian sing luwih lawas lan wis rampung karo ringkasan cendhak sing tetep **keputusan kunci lan kasunyatan saiki** lan ngilangi tangent sing wis rampung.

## Kenapa nyimpulake

- **Context window** — saben model duwe context sing winates. Riwayat dawa sing ora diedit malah bakal metu lan pesen paling lawas bakal dipapas.
- **Biaya lan latency** — sampeyan mbayar (ing token) lan ngenteni saben pesen sing sampeyan kirim maneh. Thread 50-pesen sing diproses maneh saben giliran alon lan larang; ringkasan sing ringkih murah.
- **Fokus** — back-and-forth sing ora relevan (bug sing wis sampeyan tetapake, ide sing sampeyan tolak) bisa ngalihake perhatian model saka apa sing penting saiki.

## Apa sing kudu ditahan lan apa sing kudu diilangi

```text
KEEP → decisions made ("we chose Postgres"), constraints ("must run on Node 18"),
        current state ("auth works; now wiring the dashboard"), open questions
DROP → resolved tangents, dead-end attempts, long pasted output already acted on,
        polite filler
```

## Conto

```text
# Instead of 40 messages, summarize the settled context:
Summary so far:
- Building a REST API in Express + Postgres (Node 18).
- Auth (JWT) is done and tested.
- DECISION: use Zod for request validation.
Current task: add the GET /orders endpoint with pagination.
```

IkI tetep model kanthi lengkap ing keputusan lan tugas saiki sanajan ngilangi gangguan sing dadi siji-sijine penyebab.

## Apa gunane

Nguasai kapan nyimpulake tetep asisten AI bisa nggarap sajrone context window, cepet, lan murah — lan tetep perhatian ing keputusan lan kasunyatan saiki sing tenan-tenane nyurung langkah sabanjure, tinimbang kegubungan ing riwayat sing wis rampung.