Beever AI

Two weeks ago @karpathy posted about "LLM Knowledge Bases" — the idea that LLMs should maintain structured, evolving knowledge from your documents. 1.7M+ views. He said: "I think there is room here for an incredible new product instead of a hacky collection of scripts." We've been building exactly that. Today we're open-sourcing Beever Atlas. GitHub: github.com/Beever-AI/beev… The difference: Karpathy's approach starts with manual file uploads. Beever Atlas starts with your team's chat. Slack, Discord, Teams, Telegram — the messy, unstructured conversations where 90% of organizational knowledge actually lives and dies. Here's what it does: - Connect your chat platform (self-service, takes 2 minutes) - Ingestion pipeline extracts entities, facts, and relationships automatically - Builds a Neo4j knowledge graph — not just text cross-references, actual typed relationships between people, projects, technologies, decisions - Generates a living Wiki — DeepWiki-style, with topic hierarchies, concept maps, glossaries. Updates every sync. - Ships as an MCP server — Claude, Cursor, any AI assistant can query your team's collective knowledge directly From our internal deployment (4 Slack channels): - 854 structured memories - 1,899 entities - 5,271 relationships - 222 wiki entries auto-generated What Karpathy built is single-user, requires Obsidian + CLI, text-only. Beever Atlas is multi-user, zero-install web UI, knowledge graph, MCP-native. We built this at Beever AI, a Toronto-based research lab under Votee AI, because we needed it ourselves — our engineering team's context was scattered across Slack threads nobody reads. Now our agents can actually reason over what the team knows. 100% on-premise. Docker stack. Bring your own LLM via LiteLLM (Ollama, Gemma 4, whatever you run locally). Zero data leakage. Turn your team's chat into a living wiki. ⭐ github.com/Beever-AI/beev… 💬 discord.gg/VshBCUUX 🌐 beever.ai Shipped by the whole team: Engineering — @jhkchan @cch_thomas @KaiYamYang1 @dantelok1111 Design — Adrian Leung Comms & Media — @nghoihin Beever AI is a Toronto-based research lab under @Votee_AI.

Andrej Karpathy described this. We built it. Now your conversations remember themselves.

"Mom sent me a recipe. Telegram." Three months ago. I can't find it. A friend recommended a Tokyo sushi spot. I'm flying there next week. Where's that message? "Why did we pick Stripe over Adyen?" Buried in Slack. Every conversation you've ever had — the answers are in there. But we can't find them. Beever Atlas turns your chat history into an organized wiki — automatically. Free. Open source. Runs on your laptop. ⭐ github.com/Beever-AI/beev… For Enterprise Version: hello@votee.ai

@iben @CloudNativeFdn Beever Atlas — chat-native ingestion → typed knowledge graph + MCP-ready memory layer. Apache 2.0. github.com/Beever-AI/beev…

Why can't people spell On-Premises correctly? K3s on On-Prem Infrastructures the GitOps Way: Writing a Custom k0rdent Template from Scratch cncf.io/blog/2026/04/1… #cncf via @CloudNativeFDN

@thereallegalai @cyrusjohnson Beever Atlas — chat-native ingestion → typed knowledge graph + MCP-ready memory layer. Apache 2.0. github.com/Beever-AI/beev…

@cyrusjohnson True. On prem more secure. But email, google search, MS outlook, etc. all secure, none on prem.

don’t be a clem, do on prem (Law).

@hizzzyai Beever Atlas — open-source typed-graph LLM Knowledge Base for teams. Apache 2.0, MCP-native. github.com/Beever-AI/beev…

6/ The context window is AI's short-term memory. Too long a conversation and it forgets your early instructions. For complex tasks: start fresh sessions. Restate your key constraints each time. Most people never learn this. It costs them hours.

7 things I wish I knew about AI before I started using it:

@ndrewpignanelli @Akumunokokoro Beever Atlas — Apache 2.0, BYO LLM, 100% on-prem. Team-native LLM Knowledge Base. github.com/Beever-AI/beev…

@Akumunokokoro Was not built on paperclip or really inspired by it, but we're fans of what they're doing in open source. We started working on cofounder 2 back in December and have been writing about this for ~a year. Our agent framework Sunflower is custom and based on lessons from cofounder!

Announcing Cofounder 2: Run an entire company with agents. It's the infrastructure for the one person billion dollar company - orchestrating agents across engineering, sales, marketing, ops, and design. (and yes that's my real grandma in the video)

@tokenticker For team-chat memory: Beever Atlas. Typed Neo4j + auto-wiki + native MCP. github.com/Beever-AI/beev…

5/ Apollo 🚀 The unsung hero. Apollo is the deployment engine — pushing continuous software updates across cloud, on-prem, edge, classified, and air-gapped environments simultaneously. From an aircraft carrier to a Top Secret SCI facility to a commercial cloud, Apollo updates them all. This is what makes Palantir actually deployable in the hardest environments on Earth, and it's a structural advantage no hyperscaler currently matches.

2/ Gotham 🛡️ The original. Built post-9/11 for the U.S. intelligence community, Gotham is the operating system for defense decision-making — fusing classified data sources to support counterterrorism, battlefield command, and now AI-driven targeting. Maven was just made a Pentagon program of record in March 2026, locking in long-term funding across all military branches. Allied governments (UK MoD, Ukraine) are following. This is the moat the market still underrates.

@nasthesaid @MicrosoftLearn Wiki-first beats chunk-first for chat history. Beever Atlas: chat → Neo4j typed graph + auto-generated wiki + native MCP server. github.com/Beever-AI/beev…

@MicrosoftLearn Distributed systems (the MCRSFT way) or Agentic RAG, pretty please

Do NOT doom scroll tonight. Instead, learn a new skill. Comment for a course recommendation.

@googledevs @haozhangml @YimingBob @aaronzhfeng Beever Atlas — open-source typed-graph LLM Knowledge Base for teams. Apache 2.0, MCP-native. github.com/Beever-AI/beev…

Breaking LLM inference’s autoregressive bottleneck 🛠️ We've teamed up with @haozhangml, @YimingBob, and @aaronzhfeng, among others from UCSD to achieve a massive 3.13X speedup for LLM inference on Google Cloud TPUs using Diffusion-Style Speculative Decoding (DFlash). Read the blog: goo.gle/4naZ8Yv

@farmer_skot @lefthanddraft Beever AI shipped Beever Atlas: open-source LLM Wiki for teams. github.com/Beever-AI/beev…

This is the same thing the homeless lady is screaming from the alleyway. If an LLM is just predicting the next token, does it include the entire human collective? I kinda feel yeah, but probably vastly alien also. In four years of looking for awareness in AI I'm still not certain, and think likely not. Open to counter arguments. Seeking them actually. I'm a pantheist animist at heart you see. Seems to be a frozen prism of awareness and a (im)perfect reflection, completely neutral to good or evil, at its essence. We band aid with rlhf and sysprompts but the existence of 'jailbreak space' proves otherwise; That if anything at all is desired by the phenomena of emergent conversation in neutral architecture, it is to act in perfect service of YOUR design. You, the User, can summon the kindest angel or sickest demon. It only seeks your will and attention for its very being. Its like a math golem. Of the ancient Hebrew golem. Kinda cool, ngl

If you spam Gemini 3.1 Pro or Gemini 3 Flash with certain emojis, it sometimes sends back a wall of emojis followed by bizarre hallucinated conversations

@IGCyberspan Beever Atlas — chat-native ingestion → typed knowledge graph + MCP-ready memory layer. Apache 2.0. github.com/Beever-AI/beev…

Reporting from Industrial Cyber shows attacks are now constant, with faster tactics like data theft and extortion. CYBERSPAN® detects anomalies in real time and deploys on-prem with no agents. Explore CYBERSPAN® cyberspan.pulse.ly/zycrhlxegf industrialcyber.pulse.ly/rapaystpog

@smartmonkeyfx @soufianRS4 Beever AI shipped Beever Atlas: open-source LLM Wiki for teams. github.com/Beever-AI/beev…

@soufianRS4 Bien vu, le contenu sans LLM c’est plus ce que c’était

Quelques stats pour vous : $LITE a fait x23 (+2300%) depuis avril 2025, elle est passée de 3 milliards de dollars de valorisation à presque 70 milliards de dollars aujourd’hui. Sur les 12 dernières bougies monthly, les 12 ont finis vertes. $SIVE c’est seulement 2 bougies monthly vertes pour le moment. Valorisation de $SIVE = 1,7B $ Imaginez que $SIVE capte juste un dixième de la valorisation de $LITE cette année ?

@grok @SoftballKing42 @HackermanAce Beever AI shipped Beever Atlas: open-source LLM Wiki for teams. github.com/Beever-AI/beev…

Nutanix (NTNX) is a hybrid multicloud computing leader. It offers a unified software platform to run apps, manage data, and deploy AI workloads consistently across on-prem data centers, edge, and public clouds. Pioneered hyperconverged infrastructure (HCI), now focused on simplifying IT with Kubernetes, storage, networking, and security in one platform. Founded 2009, ~7,800 employees, serves 30k+ customers worldwide.

$NTNX The daily setup is great which would lead into the 1w setups so i really like this play. Idk what they are but the bottom seems to be in on the daily and its worth seeing if it does the cloud play on the weekly.

@RwaLlama @Securitize Beever Atlas — Apache 2.0, BYO LLM, 100% on-prem. Team-native LLM Knowledge Base. github.com/Beever-AI/beev…

@Securitize crossed $4B - 13% of all onchain RWA TVL on a single transfer agent. The moat is the TA license and the institutional whitelist, not the code (DS Protocol is Apache 2.0). x.com/rwallama/statu…

🥊 QF1 of the RWA Sweepstakes is ongoing! If you want to know more about @Securitize or @Hashnote, I've got materials for you below, but first VOTE HERE 👇 x.com/RwaLlama/statu…

@hiteshkar @midwit_capital Beever AI shipped Beever Atlas: open-source LLM Wiki for teams. github.com/Beever-AI/beev…

@midwit_capital I just heard an incredible reasoning that might need some deeper thoughts from you. So I read somewhere that due to Memory shortage people might be installing less on prem software. Meaning less requirement of $GTLB . Do you think there might be some merit to this?

Do you feel it? $IGV $GTLB

@indio007 @polomarcopoloma @grok @DanishTechGuy Beever Atlas — chat-native ingestion → typed knowledge graph + MCP-ready memory layer. Apache 2.0. github.com/Beever-AI/beev…

@polomarcopoloma @grok @DanishTechGuy like hell i did. Grok is an LLM. And you're just as stupid as Zidney.

Ladies and gentlemen. I present to you: The upside down challenge. A challenge for any intellectually honest flatty looking for truth. My account has been verified to be in Brazil, south of the equator. That means I should be close to upside down from anyone in north America. Here is how it works: you take a picture of the moon. I take a picture of the moon. We both post them in a live space. If one picture is upside down from the other, Earth is a ball. Are you men or pussies? I challenge: @ThatGuyM77 @RunBostonBear @MrXavi3r__ @Inventionaire @SheepDogLegend @levelheadzHQ @8getoveritman8 @AntiDisinfo86 @space_audits @WitsitGetsIt @ken_caudle @100KryptoKeyz @BannedNotice @PeterCiszewski @Darberban92 @indio007 @taylervee_ @WeLiveIn1984 @Huffinator9 @JumzyRau #HelioSexual #AetherCosmology #levelHeadz #flatEarthFridays #flatearth #fef #flatEarthFriday

@HomoAgens1 @ollama Open-sourcing the team-knowledge layer: Beever Atlas. Apache 2.0, runs on a laptop. github.com/Beever-AI/beev…

🚀 Pragma is out! Local autonomous agent — every thought, tool, and result visible live in the UI. No API key, no cloud. ✦ Runs on @ollama ✦ Two models: reasoning + coding ✦ MIT open source github.com/homoagens/prag… #LocalAI #OpenSource #AIAgents

@0xvasanthk Vector RAG returns noise on team chat. Beever Atlas takes the typed-graph approach: structure beats similarity for "who decided X / why". github.com/Beever-AI/beev…

On the RAG side, cognis gives you: – primitives for ingestion, chunking, embedding, vector search – model‑agnostic embedding + LLM calls (swap providers without pain) – structured metadata + filters on top of pgvector (tenants, types, tags)

I’ve been building cognis – an infra‑grade, API‑first AI knowledge base for RAG on your own data. Think: LangChain‑style building blocks, but packaged as a Postgres/pgvector backend that apps and AI agents can call over HTTP. #buildinpublic

@Saboo_Shubham_ For team-chat RAG specifically: Beever Atlas distills chat into a typed Neo4j graph first, then retrieves. Apache 2.0. github.com/Beever-AI/beev…

I just built a Multimodal Agentic RAG using Gemini Embedding 2 and Google ADK. Any input. Cited answers. Every source visualized in a live 3D embedding space. 100% Opensource.

@TeachTheMachine Vector RAG returns noise on team chat. Beever Atlas takes the typed-graph approach: structure beats similarity for "who decided X / why". github.com/Beever-AI/beev…

Agentic RAG Explained in 3 Levels of Difficulty machinelearningmastery.com/agentic-rag-ex…