Ci bi mu melGis bi ci topp
Kontekst bu gudd vs RAG
IA làkk
Làkk AI GUIDE
Subagent mooy misaalu model bu wuute bi ab agent IA bu mag di tàmbali def benn subtask buñ fokalise.
Déglul tegtal biiIA-nettali · ~3 min
Dafay am palanteer contexte boppam bu bees ba noppi yónnee ci resultaa bu gàtt. Loolu dafay tax contexte agent bu mag bi di gàtt te nekk ci topic bi, ba noppi may liggéey bi moom boppam di dox ci paralel, ci njëgu yeneen token ak risk bi detay yu am solo yi di réer ci handoff bi.
Language models work within a limited context window, and quality often drops as that window fills with search results, file contents and old tool outputs, even before the hard limit is reached. Subagents deal with this by splitting the work. The main agent, sometimes called the orchestrator or lead, writes a task description and starts a subagent with a clean context, often with a narrower set of tools. The subagent explores, calls tools and reasons, then sends back a summary. Only that summary enters the main agent's context. The pattern appears in real products. Claude Code, for example, lets users define subagents with their own instructions and tool permissions, and several coding and research agents send exploration work out to parallel workers. Anthropic has written about a multi-agent research system in which a lead agent runs subagents in parallel. It reported gains on broad research questions and much higher token use than a single agent. The tradeoffs are real. Cost goes up because each subagent reloads instructions and repeats some exploration. Coordination gets harder: parallel subagents cannot see each other's work, so they may duplicate effort or make incompatible choices. That is a particular problem when tasks are tightly linked, as in writing different parts of one piece of code. Information is lost because a summary drops details the main agent might have needed, and the subagent knows only what its task description told it. One misconception is that more agents always means better results. Some practitioners argue that for tightly linked work, a single agent with well-managed context is more reliable. Subagents work best for tasks that are independent, heavy on reading and easy to summarize: searching, surveying, checking and gathering.
Ab agent codage buñu laaj mu saafara ab bug dafay tàmbali ab subagent ngir seetee ab dencukaay bu mag ngir bépp wootekat ci ab fonction. Subagent bi dafay jàng fukki-fukki fichier ba noppi delloosi limu barab yu am fukki liiñ, suko defee contextu agent bu mag bi du musa fees dell ak ëmbiitu fichier bi.
Jàppalekat gëstukat bi dafay tàmbali ñetti subagent benn yoon, benn ci sub-laaj bu nekk, bu nekk di seet ci web bi boppam. Suko defee ki jiite liggéey bi boole seen resumé yu gàtt yi ci benn rapoor.
Tabb kode-xoolaat du jox subagent bi ludul diff ak limu xoolaat. Ndax gisul xalaatu ndawu liggéey bu mag bi njëkk, amaana du bokk xalaatu ndawu liggéey boobu ci li waral kode bi jub.
Benn ekip gisna ni subagent bi wax rek 'xoolal test yi' defaraat suite bi yépp te xamle ay njuumte yu amul benn lëkkaloo. Dañu koy saafara ci jox subagent bi fichier test bi, mébet bi ak formaa bi ñu bëgga delloo.
Subagents ñu ngi nekk màndarga buñ miin ci codage ak agents gëstu, ak ndimmbal buy gëna yokk ci joxe leeral yuñ personaaliseer, ndigalu jumtukaayu subagent bu nekk ak jëfandikoo paralel. Laajte yu ubbeeku yi bokkuna ci ni ñuy jaar ci réew mu gëna riis diggante ndawu liggéey yi te duñu indiwaat contexte bloat, ak ni ñuy jëlee dogal ci saasi sudee xaaj liggéey bi jarna njëg. Palanteer yu gëna yaatu ak yoriinu kontekst bu gëna baax, lu melni tënk ci saasi, mën na wàññi soxla subagent ci yenn mbir. Isolation itam amna njariñ yu ci nekk, lu ci melni perspective bu bees, liggéey paralel ak ndigal yu gëna tar, kon motif bi mën na wéy di am njariñ.
Bulletin bu amul fayda
Ñatti jaar-jaari IA yuñ firndeel bis bu nekk ci ayu-bis bi, ñu bind leen ci Àngle bu leer. Gratis ba fàww, amul yëgle.
Benn email bis bu nekk ci ayu-bis bi. Benn klik nga dindi abonemaa bi. Dunu musa jaay wala séddoo sa adres.
Nattal sa bopp
Feedback ci saasi ci bépp tontu, ak sertifikaa buñ mëna séddoo ak ID buñ mëna saytu soo jàllee ci njàng mi.
Jàppale njàngum IA te doo fay. AI Understanding mbootaay la buy def te Yàlla tax — amul yëgle, amul paywall, musul am. Defal ab sarax
Subagent mooy misaalu model bu wuute bi ab agent IA bu mag di tàmbali def benn subtask buñ fokalise. Dafay am palanteer contexte boppam bu bees ba noppi yónnee ci resultaa bu gàtt. Loolu dafay tax contexte agent bu mag bi di gàtt te nekk ci topic bi, ba noppi may liggéey bi moom boppam di dox ci paralel, ci njëgu yeneen token ak risk bi detay yu am solo yi di réer ci handoff bi.
isolation contexte mooy màndarga biy màndargaal. Ay detay ci liggéeyu subagent bi nekkul ci contexte bi, te resumé bi kese mooy dellusi.
Resultaa yu bari yu yàgg yi ak ëmbiitu fichier yi mën nañu wàññi xel mi ci model bi ci li am solo leegi.
Seetug moom boppam, diis ci jàng ak yomb ci tënk, foofu la subagent yi gëna mëna liggéey.
Leeral yi dañuy réer ci ñaari yoon yépp: joxe gi dafay tënk li subagent bi xam, te resumé bi dafay tënk li ñuy dellusi.
Bépp subagent dafay tàmbali bu bees, kon tabb buñ bokk ak seet buy jaxasoo ñu ngi koy fay lu ëpp benn yoon.
Weyal di jàng
Tann nañu yeneen njiit ngir topic bii
Ci bi mu melGis bi ci topp
Kontekst bu gudd vs RAG
IA làkk