Matching posts

Matching terms are highlighted

Prompt has Grok Bot scan your history and build your bot team

A prompt tells Grok Bot to scan chats, email, calendar and posts, rank recurring manual tasks by hours lost, then create a scheduled bot for the top three and run each one's first task.

personal-agent@jumperz　Filed in No.012

3 days agoCopy original

Opus 5.5 renders a Hangzhou history film frame by frame in code

…容易啊，由 Opus 逐帧写代码渲染，改了三四版，跑了近6个小时才出来 从良渚到现在，五千多年的历史压缩进4分51秒 西湖是杭州的来处，钱塘江是杭州的去处

claude@wangray　Filed in No.010

5 days agoCopy original

Grok Bot voice call demo: interruptible, searches past chats, drafts transcript

A demo shows Grok Bot voice calls letting you interrupt and be heard, searching past chats mid-call, and dropping a transcript plus draft into chat for approval via the Action Button.

grok@GrokBotRadar　Filed in No.012

6 days agoCopy original

POST

Grok Bot can agentically crawl an entire profile's history

The author says regular Grok reads only the latest 150 posts and 300 replies, while Grok Bot agentically crawls a profile back to its creation.

grok@anonitext　Filed in No.012

7 days agoCopy original

Berkeley-led paper resets auto-research AI history, tested on DeepSeek V4.1 Flash

A preprint keeps only the best code and scores while wiping agent history; with four DeepSeek V4.1 Flash agents it beat independent runs by ~15% and a history-keeping single agent by ~39% across four tasks.

deepseek-v4.1@LangChainJP　Filed in No.020

8 days agoCopy original

POST

Codex desktop app: edit earlier messages and continue from that point

Codex's desktop app now lets users edit an earlier message in a thread and resume from there, dubbed time travel; whether file changes roll back is unconfirmed.

codex@TokenGremlin　Filed in No.004

9 days agoCopy original

POST

Codex desktop adds message editing on day 15 of 28

OpenAI Codex's 28-day run reached day 15 with message editing in desktop threads, while the poster tallies only two real wins and five resets in the first 14 days.

codex@ivanainai　Filed in No.004

9 days agoCopy original

Scoring history with Jev and gating actions on confidence

The author uses Jev to score history entries into four tiers, cutting 1M tokens to 86K, routes low-confidence questions to Kimi K3, and keeps a fixed 0.95 publish threshold while recalibrating the 0.85 line weekly.

jev@DonatelloXPro　Filed in No.006

11 days agoCopy original

Grok Bot context strategy: stop feeding it all your history

The author calculates that sending 40,000 tokens of history per call adds up to 40 billion tokens over a million calls, and argues for retrieving only what the current request needs before handing it to Grok.

personal-agent@kocer_eth　Filed in No.012

14 days agoCopy original

Kobra ships a chat history component

Kobra released an embeddable chat history component that lifts history out of messages with the agent and takes you back to the original spot.

agent@haaarshsingh　Filed in No.004

15 days agoCopy original

Showing10of60posts

10 per batch

Keep scrolling to reveal posts 11–20

Showing the first 60 matches. Narrow the query with a model name or tag to find more specific results.