Memory

Evidence (5)

• 5Anthropic News·7/9/2026product_launch
  Anthropic Adds a Claude Usage Reflection Dashboard

  Anthropic has launched a beta reflection dashboard that lets users review how they use Claude over time, including usage patterns, task types, and prompts to consider whether AI is serving their goals. The feature also supports quiet hours and break reminders, and excludes incognito chats and certain sensitive data sources from the insights.

• 5Hacker News·7/3/2026research
  PostgreSQL and the OOM Killer: Why Ubicloud Uses Strict Memory Overcommit

  Ubicloud explains why its managed PostgreSQL service keeps Linux strict memory overcommit enabled to reduce the risk of catastrophic out-of-memory kills. The post describes how the setting helps fail early, the impact of a kernel bug that temporarily forced the team to disable it, and the heuristic used to choose a commit limit.

• 4OpenAI Blog·4/10/2026product_launch
  Personalize ChatGPT with Instructions and Memory

  OpenAI explains how users can tailor ChatGPT with custom instructions and memory so responses better match their preferences and context. The feature aims to make interactions more consistent and relevant across chats.

• 6OpenAI Blog·1/29/2026research
  OpenAI details its in-house data agent

  OpenAI describes an internal AI data agent built to reason over large datasets and produce reliable insights in minutes. The system uses GPT-5, Codex, and memory to support faster analysis workflows inside the company.

• 5OpenAI Blog·1/7/2026product_launch
  Tolan builds a voice-first AI companion with GPT-5.1

  Tolan uses GPT-5.1 to power a voice-first AI companion designed for natural, low-latency conversation. The system combines real-time context reconstruction with memory-driven personalities to make interactions feel more coherent and personal.