Even after working with AI for a long time, I kept having to explain what I had already done. Switching models meant starting the same story again. That frustration led me to create VALE: a place where shared memories stay together, whichever AI I use. As those memories grow, I want Life Galaxy to help me look back on my interests and the life I have lived.
AI keeps advancing. The memories you build together should carry forward as time passes and models change. VALE is a personal memory engine and workspace that keeps your original conversations available throughout your life.
Connect from a phone, tablet or another computer →7 min 11 s · Features and workflows shown in the VALE app. Game scenes and lifetime changes are illustrative recreations.
Find overlooked steps and decisions in the original conversation and use them in today’s answer.
Select relevant original passages to reduce unnecessary input tokens and help the answer focus on important evidence.
Switch models or work in several chats at once, sharing the memory of the same universe.
“I’m writing a personal statement. Help me organise my past projects and work.”
In the video, VALE finds the app you shipped, the bug you fixed, and the checks you ran to preserve existing notes. Steps you had forgotten become useful experience for the next task.
The original records who said what and in which context. That evidence helps distinguish your actions from reference examples and plans you did not pursue, reducing confusion between different stories.
More context does not automatically mean a better answer. Repeatedly including unrelated conversation uses more input tokens and can bury important evidence, reducing answer quality. What matters is supplying the context the current task actually needs.
AI reads the context supplied for each answer. When you ask for a personal statement, old travel plans and unrelated small talk can end up in that input too. This adds input tokens and can make the relevant experience and its conditions harder to use.
More input tokens · Less focused answers
Some AI apps automatically summarise or compact earlier context to keep long conversations going. The conversation continues, but it can be hard to tell which conditions disappeared from the context used for an answer, and when. Without the original, AI can answer confidently while missing what the summary left out.
Gaps you notice in an answer—or never notice at all
You may keep a document summarising the design or progress and reattach it along the way. It helps work continue, but cannot restore details it omitted. Reattaching the same document puts that content back into the input, and someone still has to maintain and supply it.
Missing details remain · Repeated input and upkeep
VALE preserves the original conversations and selects relevant memories for the current question. Short headers and connections between memories lead back to the source. The wording and context behind a decision stay available without putting the entire archive into every answer.
“I won the competition” is different from “This is a sample personal statement, not my experience.” The original records who said what and in which context. VALE brings that evidence into the answer to help keep the two apart.
Short chats may rarely run into long-context compaction. But a new chat or a different model can still mean explaining past decisions and experience again. In VALE, chats and models share the memory of the same universe. Memory helps you return to work spread across short conversations as well as continue long ones.
Anthropic describes context compaction for long-running tasks and notes that aggressive compaction can lose details whose importance emerges later. Lost in the Middle shows that a model’s ability to use information can vary with its position in a long input. These findings help explain why storing more memory and reading more context are different decisions.
Start with Claude and switch to GPT in the same chat. Or open three chats: Claude checks code, GPT writes an introduction, and Grok prepares review questions.
The three models share the memory of the same universe. Connect newer models for answers while carrying your accumulated memories forward in one VALE workspace.
VALE mode lets you enter a universe, assign work, and see progress and results in one place. You choose which universe to access; each universe’s memory stays independent.
Arrange mail, calendars, market news and coding progress as cards. Collapse, resize or rearrange them around the task at hand. Assign several jobs to workbenches, receive their results together, and add follow-up instructions while work continues.
Choose which of your connected AI models will join the meeting. Each reads and responds to the other models’ contributions. You can also assign each participant a role, such as product planner, engineering lead or security reviewer.
For example: “We want to pilot a customer support platform for businesses in four weeks. Decide what must ship, what can wait and what needs checking before release.” Select Claude, GPT and Grok, then assign them product, engineering and security roles.
When the product model proposes automatic classification, the engineering model narrows the scope to fit the schedule. The security model adds tenant isolation and authorization checks as release conditions. Their proposals and objections help identify priorities and risks to check. You can adjust roles or direction along the way, and keep the summary and minutes.
You choose the models and their roles. This combination is an example, shown in the actual app with demonstration dialogue.
“The customer support dashboard still shows resolved tickets when I select Open. Help me fix it.” AI remotely controls the browser, changing status and date filters and clicking Apply to reproduce the issue. Review the same screen together, then verify the same conditions after the fix.
At the same time, three agents handle the ticket filter bug, query performance improvements and filter regression tests. Keep the conversation going while they work, and receive their results in the same chat.
Browser tasks require Google Chrome.
Three agent tasks on the left; the dashboard filtered to Open tickets from the last seven days on the right. Actual app UI with demonstration data.
Open the small orb in live conversation mode and talk to VALE as you work. Ask about game strategy, assign coding tasks by voice, and hear the results while you keep playing.
When you want to focus with the keyboard, choose a universe in the chat menu. Attach files, review material, write and develop in a familiar workspace.
A universe is an independent space for conversations and memories. Each memory becomes a star; connected memories appear as galaxies. The time dial reveals stars in the order they were recorded, and selecting a galaxy highlights related memories.
Over time, you can look back at what mattered to you: perhaps a career-centred universe in your twenties gives way to a growing family galaxy later in life. Reopen the conversations years later to revisit the concerns and choices of that time.
The lifetime changes shown in the video illustrate long-term use.
An arc is a story that unfolds over time. Follow a graduation project, a personal app or an internship from its beginning through setbacks and results.
An entity is a person, tool, project or other subject mentioned in a conversation. Select your mother to gather records about walks and trips across different topics. Open a short header to read the full exchange between you and VALE.
The structure map follows connections between galaxies, galaxy clusters and superclusters, from broad topics to smaller groups.
Keep your working self and writing self in separate universes. Universes do not share memory, so each conversation can stay in its own context.
A black hole is a space for private thoughts within the same universe. It can draw on general memory, but its contents do not spread to conversations outside. You can open the records yourself.
VALE is a local app with no central server. The memory database is stored only on your computer, where the coordinates used to find memories are computed too.
Search records, filter by date, restore deleted records from the trash and back up your records.
Rediscover forgotten experience, save tokens by reading the originals you need, and work with different AIs that share your memory.
The longer you use it, the more of your life you can look back on. VALE, an AI workspace that remembers you.
VALE runs on your own computer. The Mac installer is signed and notarized by Apple; the app tells you when a new version is out.
View releases on GitHub ↗