🐦 Twitter Post Details

Viewing enriched Twitter post

@dair_ai

Autonomous Memory Management in LLM Agents LLM agents struggle with long-horizon tasks due to context bloat. As interaction history grows, computational costs explode, latency increases, and reasoning degrades from distraction by irrelevant past errors. The standard approach is append-only: every thought, tool call, and response permanently accumulates. This works for short tasks but guarantees failure for complex exploration. This research introduces Focus, an agent-centric architecture inspired by slime mold (Physarum polycephalum). The biological insight: organisms do not retain perfect records of every movement through a maze. They retain the learned map. Focus gives agents two new primitives: start_focus and complete_focus. The agent autonomously decides when to consolidate learnings into a persistent Knowledge block and actively prunes the raw interaction history. No external timers or heuristics forcing compression. It declares what you are investigating, explores using standard tools, and then consolidates by summarizing what was attempted, what was learned, and the outcome. The system appends this to a persistent Knowledge block and deletes everything between the checkpoint and the current step. This converts monotonically increasing context into a sawtooth pattern: growth during exploration, collapse during consolidation. Evaluation on SWE-bench Lite with Claude Haiku 4.5 shows Focus achieves 22.7% token reduction (14.9M to 11.5M tokens) while maintaining identical accuracy (60% for both baseline and Focus). Individual instances showed savings up to 57%. Aggressive prompting matters. Passive prompting yielded only 6% savings. Explicit instructions to compress every 10-15 tool calls, with system reminders, increased compressions from 2.0 to 6.0 per task. Capable models can autonomously self-regulate their context when given appropriate tools and prompting, opening pathways for cost-aware agentic systems without sacrificing task performance. Paper: https://t.co/bVkeQlrvGJ Learn to build effective AI agents in our academy: https://t.co/zQXQt0PMbG

Media 1
View on Twitter

📊 Media Metadata

{
  "media": [
    {
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2011806092737827206/media_0.jpg?",
      "media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2011806092737827206/media_0.jpg?",
      "type": "photo",
      "filename": "media_0.jpg"
    }
  ],
  "processed_at": "2026-01-18T17:31:23.307000",
  "pipeline_version": "2.0"
}

🔧 Raw API Response

{
  "type": "tweet",
  "id": "2011806092737827206",
  "url": "https://x.com/dair_ai/status/2011806092737827206",
  "twitterUrl": "https://twitter.com/dair_ai/status/2011806092737827206",
  "text": "Autonomous Memory Management in LLM Agents\n\nLLM agents struggle with long-horizon tasks due to context bloat.\n\nAs interaction history grows, computational costs explode, latency increases, and reasoning degrades from distraction by irrelevant past errors.\n\nThe standard approach is append-only: every thought, tool call, and response permanently accumulates. This works for short tasks but guarantees failure for complex exploration.\n\nThis research introduces Focus, an agent-centric architecture inspired by slime mold (Physarum polycephalum). The biological insight: organisms do not retain perfect records of every movement through a maze. They retain the learned map.\n\nFocus gives agents two new primitives: start_focus and complete_focus. The agent autonomously decides when to consolidate learnings into a persistent Knowledge block and actively prunes the raw interaction history. No external timers or heuristics forcing compression.\n\nIt declares what you are investigating, explores using standard tools, and then consolidates by summarizing what was attempted, what was learned, and the outcome. The system appends this to a persistent Knowledge block and deletes everything between the checkpoint and the current step.\n\nThis converts monotonically increasing context into a sawtooth pattern: growth during exploration, collapse during consolidation.\n\nEvaluation on SWE-bench Lite with Claude Haiku 4.5 shows Focus achieves 22.7% token reduction (14.9M to 11.5M tokens) while maintaining identical accuracy (60% for both baseline and Focus). Individual instances showed savings up to 57%.\n\nAggressive prompting matters. Passive prompting yielded only 6% savings. Explicit instructions to compress every 10-15 tool calls, with system reminders, increased compressions from 2.0 to 6.0 per task.\n\nCapable models can autonomously self-regulate their context when given appropriate tools and prompting, opening pathways for cost-aware agentic systems without sacrificing task performance.\n\nPaper: https://t.co/bVkeQlrvGJ\n\nLearn to build effective AI agents in our academy: https://t.co/zQXQt0PMbG",
  "source": "Twitter for iPhone",
  "retweetCount": 50,
  "replyCount": 3,
  "likeCount": 329,
  "quoteCount": 4,
  "viewCount": 38975,
  "createdAt": "Thu Jan 15 14:22:04 +0000 2026",
  "lang": "en",
  "bookmarkCount": 299,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2011806092737827206",
  "displayTextRange": [
    0,
    303
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "dair_ai",
    "url": "https://x.com/dair_ai",
    "twitterUrl": "https://twitter.com/dair_ai",
    "id": "889050642903293953",
    "name": "DAIR.AI",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/1643277398522187778/31dedbLo_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/889050642903293953/1742055232",
    "description": "",
    "location": "",
    "followers": 85514,
    "following": 1,
    "status": "",
    "canDm": true,
    "canMediaTag": true,
    "createdAt": "Sun Jul 23 09:12:45 +0000 2017",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 4016,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 122,
    "statusesCount": 2821,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "2012903315890225220"
    ],
    "profile_bio": {
      "description": "Democratizing AI research, education, and technologies. New Claude Code cohort: https://t.co/XCpQTjR9hg",
      "entities": {
        "description": {
          "urls": [
            {
              "display_url": "dair-ai.thinkific.com/courses/claude…",
              "expanded_url": "https://dair-ai.thinkific.com/courses/claude-code-for-everyone-cohort-3",
              "indices": [
                80,
                103
              ],
              "url": "https://t.co/XCpQTjR9hg"
            }
          ]
        },
        "url": {
          "urls": [
            {
              "display_url": "dair.ai",
              "expanded_url": "https://www.dair.ai/",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/lkqPZtMmfU"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "display_url": "pic.twitter.com/H2AH4t2isS",
        "expanded_url": "https://twitter.com/dair_ai/status/2011806092737827206/photo/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "features": {
          "large": {
            "faces": [
              {
                "h": 94,
                "w": 94,
                "x": 796,
                "y": 596
              },
              {
                "h": 92,
                "w": 92,
                "x": 782,
                "y": 1010
              },
              {
                "h": 292,
                "w": 292,
                "x": 603,
                "y": 24
              }
            ]
          },
          "orig": {
            "faces": [
              {
                "h": 94,
                "w": 94,
                "x": 796,
                "y": 596
              },
              {
                "h": 92,
                "w": 92,
                "x": 782,
                "y": 1010
              },
              {
                "h": 292,
                "w": 292,
                "x": 603,
                "y": 24
              }
            ]
          }
        },
        "id_str": "2011806088761655301",
        "indices": [
          304,
          327
        ],
        "media_key": "3_2011806088761655301",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARvrXvqNm0AFCgACG+te+3qa0YYAAA==",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAABCgABG+te+o2bQAUKAAIb6177eprRhgAA",
            "media_key": "3_2011806088761655301"
          }
        },
        "media_url_https": "https://pbs.twimg.com/media/G-te-o2bQAUlwO3.png",
        "original_info": {
          "focus_rects": [
            {
              "h": 886,
              "w": 1582,
              "x": 0,
              "y": 0
            },
            {
              "h": 1582,
              "w": 1582,
              "x": 0,
              "y": 0
            },
            {
              "h": 1796,
              "w": 1575,
              "x": 0,
              "y": 0
            },
            {
              "h": 1796,
              "w": 898,
              "x": 224,
              "y": 0
            },
            {
              "h": 1796,
              "w": 1582,
              "x": 0,
              "y": 0
            }
          ],
          "height": 1796,
          "width": 1582
        },
        "sizes": {
          "large": {
            "h": 1796,
            "w": 1582
          }
        },
        "type": "photo",
        "url": "https://t.co/H2AH4t2isS"
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "urls": [
      {
        "display_url": "arxiv.org/abs/2601.07190",
        "expanded_url": "https://arxiv.org/abs/2601.07190",
        "indices": [
          2002,
          2025
        ],
        "url": "https://t.co/bVkeQlrvGJ"
      },
      {
        "display_url": "dair-ai.thinkific.com",
        "expanded_url": "https://dair-ai.thinkific.com/",
        "indices": [
          2078,
          2101
        ],
        "url": "https://t.co/zQXQt0PMbG"
      }
    ]
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "article": null
}