Emerging Builder. $MEOW @SalvorKnows @SalvorLabsAI https://xitter.cf/t.co/szs1a1vcJq https://xitter.cf/t.co/h49iBsf9BT
Antalya
Joined November 2010
- Tweets2.6K
- Following58
- Followers258
- Likes21.1K
Pinned Tweet
44 days of shipping with just two of us, and this is only the start. 🚀 Welcome to the new devs and our marketing lead, and let's go faster together. $SALVOR 🧵
1/6 - Note to all $SALVOR fam on how we operate internally. I believe transparency is important. 🧵🪡
For 44d @BlockchainDan & @nitraeX alone have been solo leveling this project as best we can. Translating between Turkish and English 24/7.
Your AI forgot again?
Salvor keeps what your project learned, as plain files in your repo. Free and open source.
Salvor MCP (alpha): one brain for Claude, ChatGPT and Codex.
In development: x402 + $SALVOR, so agents pay per use within a limit you set.
@SalvorLabsAI
nitrateX retweeted
A lesson crossed AI tools.
In the alpha demo, Claude saved a missing feature as a lesson. ChatGPT checked it, fixed the code and added a lesson. Claude reviewed and merged.
$SALVOR #AI #AIagents #MCP
@BlockchainDan @SalvorKnows @SalvorKnows @nitraeX
nitrateX retweeted
Two lessons disagree. Which one should guide the work? Salvor flags the conflict for a human decision, keeping judgment with the people responsible.
$SALVOR #AI #AIagents #OpenSource
@SalvorKnows @SalvorLabsAI @BlockchainDan
nitrateX retweeted
OpenAI Reports Covert Underperformance in GPT-6 Astra share.google/32Argsz2HRZ8cP2…
nitrateX retweeted
Don't want to miss this one! $SALVOR
📢 Schedule Update
Our AMA with @SalvorKnows has been rescheduled to Sunday, 8 PM UTC due to final coordination.
We appreciate your understanding and look forward to seeing you there.
— Whale’s Hideout
nitrateX retweeted
AMA LATER TODAY!
See you at 8p UTC in @WHALESHIDEOUT telegram!
Get ready for a live demo showing MCP dev progress over the past week, general Q&A followed by an AMA.
nitrateX retweeted
Sounds familiar.
MIT proved every major AI model is secretly converging on the same "brain."
The “platonic representation hypothesis” proposes that different AI models are moving toward similar ways of organizing information.
A vision model learns from images. A language model learns from text.
Despite those differences, researchers found that stronger models can develop more similar representations.
This does not mean their internal structures are identical. The finding concerns growing alignment in the relationships they learn between concepts.
The hypothesis draws on Plato’s idea that different observations reflect a shared underlying reality.
Images and text show different aspects of the same world. Models trained on each may therefore learn some of the same patterns.
The paper proposes that as models improve, their representations move toward a shared statistical picture of reality.
That remains a hypothesis. It does not prove that every major AI has the same “brain” or has discovered the underlying structure of the universe.
The central idea: different training paths may lead models toward increasingly similar representations.
nitrateX retweeted
Take a look at the @BagsApp deployed tokens with >$100k mcap and real proof of development. Some real gems in there 💎. Using Bags platform to support building real tech.
nitrateX retweeted
It’s time for a new challenge
Drop some ideas below 👇
bags.fm/launch
nitrateX retweeted
We agree! L1 = compressed version, L2 = deeper understanding where needed. Knowledge artifacts provide rationale and understanding. Drives better results in fewer turn. Open source, works across sessions, vendors, teams and models. Salvorknows.dev
You Don't Need Long Context
TL;DR: Long context may be unnecessary for LLM agents. By compressing history into a short semantic state, agents maintain competitive performance even when resetting context after every action, while significantly reducing inference costs.
arxiv.org/abs/2611.05306
nitrateX retweeted
Claude öğrendi, ChatGPT hatırladı. 🧠
Claude terminalde öğrendiğini projeye kaydediyor. ChatGPT aynı bilgiyi Salvor MCP (alpha) ile buluyor, kontrol ediyor ve kendi notunu ekliyor.
Bilgi modele değil, projeye ait.
@SalvorLabsAI
@SalvorKnows
@blockchaindan
@nitraeX
nitrateX retweeted
Does your AI forget you every new chat? 🐠
Same rule, same question, day after day.
@SalvorLabsAI keeps the notes in your project. A human approves what the AI learns, and every AI reads the same notes first.
The memory lives in the project.
@BlockchainDan
@SalvorKnows
@SalvorKnows
@nitraeX
nitrateX retweeted
one-off script? You do not need Salvor OS.
It fits repeated agent work in the same repo: long-lived codebases, teams and tool switchers. If there is no second session, there is nothing to salvage.
$SALVOR #AI #AIagents #OpenSource
@BlockchainDan
@SalvorKnows @SalvorLabsAI @nitraeX
nitrateX retweeted
Planned for Salvor+: agents will work in parallel on shared, audited memory. Multi-agent swarms are in development.
$SALVOR #AI #AIagents #OpenSource
@SalvorKnows @SalvorLabsAI @BlockchainDan
Your AI forgets everything when the session ends.
Salvor gives every AI one shared project memory. It lives in your repo, and you approve what gets saved.
One brain. Every AI. 🧠
@SalvorLabsAI
nitrateX retweeted
solana:4KxtNWc5XTL3cuyc8PJMchqB6RYggvEFeMub727GBAGS chart is forming an ascending trend, and a breakout above ~841K could trigger a strong upward move 🧑🍳
nitrateX retweeted
My 2 conviction bags
- solana:Hg5Ja55T5wESq4vyFoiVCMeHXtGyVA69X2UHq8hgpump meme with best comm ,first pump pair,og pumpfun name,dog wif BATON- Billy possible
- solana:4KxtNWc5XTL3cuyc8PJMchqB6RYggvEFeMub727GBAGS best n unique AI project,dev is SYS founder that reached 2B. The best tech backed by big whales. 100-200M easy imo.
Belief n patience is the key.
NFA DYOR
nitrateX retweeted
No extra account to remember.
Salvor OS uses plain files in your repo, without a hosted service. Keep secrets out of .salvor: it is shared project knowledge.
$SALVOR #AI #AIagents #OpenSource
@BlockchainDan
@SalvorKnows
@SalvorLabsAI
@nitraeX
nitrateX retweeted
Salvor: Your Repo Remembers
Git preserves what changed. Salvor preserves what the project learned.
AI coding agents are powerful — until you hit /clear or start a fresh session. Suddenly three days of hard-won context, decisions, failed approaches, and “why we did it this way” vanish. The code is still there. The understanding is gone.
Salvor (built and maintained by @BlockchainDan) fixes this with a simple but powerful idea: an open-source, repo-native knowledge layer. Knowledge lives in your repository as structured, versioned Markdown (under .salvor/), not in ephemeral agent memory.
What Salvor Actually Does
Captures decisions, domain learnings, failures, and deferred findings — all user-approved and governed
Two-tier memory: L1 (fast current state) + L2 (deeper history and reasoning)
Travels with the code — new teammates, different agents, or future sessions start with real context
Vendor-agnostic and portable across Claude, Codex, Gemini, and others
Completely open-source at its core (Salvor OS), with commercial extensions via Salvor Labs & Salvor+
No more re-teaching your agent the same lessons. No more knowledge disappearing between sessions. The project itself becomes smarter over time.
Learn more & get started:
salvorknows.dev/
#Salvor #AI #CodingAgents #OpenSource $SALVOR
@SalvorKnows
@SalvorKnows @nitraeX
🤖 Made with AI
