Video yükleniyor...

Video Yüklenemedi

Bu video yüklenirken bir sorun oluştu. Bu geçici bir ağ sorunundan kaynaklanıyor olabilir veya video kullanılamıyor olabilir.

Ana Sayfaya Dön

We've raised $6.5M to kill vector databases. Every system today retrieves context the same way: vector search that stores everything as flat embeddings and returns whatever "feels" closest. Similar, sure. Relevant? Almost never. Embeddings can’t tell a Q3 renewal clause from a Q1 termination notice if the language is... close enough. A friend of mine asked his AI about a contract last week, and it returned a detailed, perfectly crafted answer pulled from a completely different client’s file. Once you’re dealing with 10M+ documents, these mix-ups happen all the time. VectorDB accuracy goes to shit. We built HydraDB for exactly this. HydraDB builds an ontology-first context graph over your data, maps relationships between entities, understands the 'why' behind documents, and tracks how information evolves over time. So when you ask about 'Apple,' it knows you mean the company you're serving as a customer. Not the fruit. Even when a vector DB's similarity score says 0.94. More below ⬇️show more

Nishkarsh

8,943 subscribers

3,879,458 görüntüleme • 6 ay önce •via X (Twitter)

Eğitim Haberler & Politika Bilim & Teknoloji

Anya Rossi• Live Now

Private livecam show

66 Yorum

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

If your AI agents keep retrieving the wrong context, explore HydraDB: Grateful to our investors who backed this vision: @Sky9Capital, @JeffDean, @gokulr, @shyamalanadkat, @klyap_, @0xJsum, @laura_yao, @caldbeckj, @anshulbhide, and @SeanZCai. @ashishkakran, @missionstcap, @prateeks, @MartinGTobias, @PickensAllison, @wadefoster, aekyi, @PuriSid, @mattsechrest, @CryogenicPlanet

Jeff Dean profil fotoğrafı

Jeff Dean6 ay önce

Congrats!

Rohit Ghumare profil fotoğrafı

Rohit Ghumare6 ay önce

Open source version:

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

We’re expanding our team - let’s chat?

Rohit Ghumare profil fotoğrafı

Rohit Ghumare5 ay önce

DM'ed.

Nathan profil fotoğrafı

Nathan6 ay önce

> The project "strawberry" and the fruit "strawberry" are the same word, but completely different contexts. VectorDBs cannot tell the difference. Except they can? The vector db simply abstracts fast storage and similarity computation via HNSW with a convenient API. Sure if you use static word embeddings they will map and be identical, but any decent text embedding model worth its salt today will be able to differentiate between these two concepts. And that has EVERYTHING to do with the embedding model and NOTHING to do with the vector db

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

tbh its not a vectorDB fault - its a function of how embeddings are computed and stored. different things. different problems. unfortunately no one wanted to hear about embeddings in our video.

Kevin Qi profil fotoğrafı

Kevin Qi6 ay önce

Relational databases have existed since the advent of computers. Why do we need to reinvent the wheel every year?

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

RDBMS directly can't be used as context engineering infra for your AI to work with complex enterprise knowledge bases

Kevin Qi profil fotoğrafı

Kevin Qi6 ay önce

You are saying that a competent tech team cannot reproduce what you offer simply by using sql and a good pre processing system to build the same relationships you tout?

Kevin Qi profil fotoğrafı

Kevin Qi6 ay önce

And then on LLM requests, just show the returned relevant request into the context window ?

Tiffany Fong profil fotoğrafı

Tiffany Fong6 ay önce

No error message. Just looks right. You'd never know.

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

No @hydra_db = ngmi

Jason Kneen profil fotoğrafı

Jason Kneen6 ay önce

Not open-source? Nah.

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

if we open source will you become a customer?

Jason Kneen profil fotoğrafı

Jason Kneen6 ay önce

if there is value in what the SAAS offers yes. If I can develop it myself with AI because it's overpriced then no. And if you don't open-source it the community will just build an OSS one anyway. So this is your chance to do that, own the repo, guide the development AND you have a ready made recruitment pool and advocate pool of developers who will be contributing, forking and adding to your product! A simple license that insists derivatives have to be open source and you can include their features in the core and you've got it all covered

GuestHiveAI profil fotoğrafı

GuestHiveAI6 ay önce

This is exactly the problem I hit building an AI concierge for short-term rentals. Guest asks "where's the vacuum?" — pure vector search returns the cleaning fee policy. High similarity score. Completely wrong. So I stopped trusting embeddings alone. Built a hybrid system: → Semantic search (0.65) + full-text match (0.25) + recency bias (0.10) → Live context injection (who's checked in, which property, what time) → 3-level gatekeeper that only surfaces emergency info when the situation actually calls for it All in PostgreSQL. No separate vector DB. No $6.5M. The real unlock isn't better embeddings — it's knowing WHEN and WHY someone is asking.

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

we're expanding our team. let's give it a shot?

GuestHiveAI profil fotoğrafı

GuestHiveAI6 ay önce

5 weeks. Solo. 128K lines of Python. Running in production right now handling real guest conversations 24/7. If that's the energy you're hiring for — DMs open

Simspron profil fotoğrafı

Simspron6 ay önce

@contextkingceo Pretty sad when you can’t even write your own tweets

GuestHiveAI profil fotoğrafı

GuestHiveAI6 ay önce

@contextkingceo 🤖

vixhaℓ profil fotoğrafı

vixhaℓ6 ay önce

Finally someone brave enough to say it out loud: vector DBs are gaslighting our agents.

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

LETS SHOUT THIS OUT

JT profil fotoğrafı

JT6 ay önce

this is why AI search across internal tools always feels slightly off

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

100%

Subah Wadhwani profil fotoğrafı

Subah Wadhwani6 ay önce

LFG! Congrats guys. Was so much fun working with you on this.

Rohan Paul profil fotoğrafı

Rohan Paul6 ay önce

I dont want my agents to reset their brains every session. Moving away from flat vector retrieval toward a structured memory tree is truly a great practical approach for solving agent continuity.

Soham Ratnaparkhi profil fotoğrafı

Soham Ratnaparkhi6 ay önce

Flat embeddings + 10M docs = confident hallucinations at scale. The fix isn't a better embedding model. It's a fundamentally different retrieval layer - one that tracks relationships, ownership, and temporal state. Super proud to be building this since Day 0 Big year ahead 🚀 and ofc, congrats boss - lesss goo

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

yessir, can we please make @hydra_db 100x better?

Soham Ratnaparkhi profil fotoğrafı

Soham Ratnaparkhi6 ay önce

@hydra_db Yessir

Carl Vellotti 🥞 profil fotoğrafı

Carl Vellotti 🥞6 ay önce

LLM search works great on 500 docs, dies at 10M. And you never realize until it's too late.

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

LLM search at demo day: genius LLM search in prod: confidently returns your CEO's lunch order as a "relevant policy document"

CG profil fotoğrafı

CG6 ay önce

Finally someone did something about this, cool launch video too.

Charly Wargnier ♨️ profil fotoğrafı

Charly Wargnier ♨️6 ay önce

Congrats @contextkingceo! "Similar ≠ relevant" is the exact bottleneck of naive RAG rn. great to see you've built this 🤗

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

thank you @DataChaz! that bottleneck is exactly what pushed us to build this

Hannaan Kirmani profil fotoğrafı

Hannaan Kirmani6 ay önce

so you're saying it won't go looking for my ex when i ask it to find X?

dinos profil fotoğrafı

dinos6 ay önce

nice marketing video. where's the benchmark?

Vaibhav Domkundwar profil fotoğrafı

Vaibhav Domkundwar6 ay önce

A lot of Better portcos are going to love this Nish — sending your way! Congrats on the launch. 🚀

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

Thank you sooo much @vaibhavbetter! So excited to have you onboard

Raul Junco profil fotoğrafı

Raul Junco6 ay önce

Interesting idea. Vector search is great, but it struggles with context as datasets get larger. Adding a graph or ontology layer to capture relationships could help avoid a lot of those weird retrieval mistakes. Congrats!

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

exactly, and that's what we're solving for at @hydra_db!

Arlan profil fotoğrafı

Arlan6 ay önce

lowk @turbopuffer is still very good but will try out ur product soon

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

@turbopuffer we think turbopuffer is great - we solve different problems :) if you want relations, timeline of how context has evolved, deal with messy unstructured knowledge then you need to try @hydra_db

Brian Chew profil fotoğrafı

Brian Chew6 ay önce

Congrats on the launch! - x icon in footer still redirects to old handle: - site title still says cortex - the site's opengraph thing still shows cortex just wanted to share in case it helps!

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

thank you for pointing it out

Sean Cai profil fotoğrafı

Sean Cai6 ay önce

Excited to back since day one :) Excellent anti-agi hedge, but I haven't been so excited about a vector db adjacent product since turbopuffer. There are personally some inane use cases around migrating enterprises to this context graph to better collect their reasoning traces for rl envs!

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

Ha anti agi hedge I love that. Grateful for your support and counsel as always!

Mikael profil fotoğrafı

Mikael6 ay önce

Damn dude you got a lot of investors

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

grateful for their support

Average Database CEO profil fotoğrafı

Average Database CEO6 ay önce

Did you mean $60.5m?

polihedge profil fotoğrafı

polihedge6 ay önce

1. Don't want to book a demo 2. Pricing page loops to homepage 3. Several links in footer do not exist 4. Privacy Policy and Terms of Use are crossed links Lots of errors on the vibe code template site for HydraDB.

Antaripa Saha profil fotoğrafı

Antaripa Saha6 ay önce

the problem statement is very real and have faced such issues during multiple client projects, would love to test out hydradb. congratulations, Nishkarsh!

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

this is sooo real...i feel it

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

Amazing - this goes in our next launch video

Siddhartha Saxena profil fotoğrafı

Siddhartha Saxena6 ay önce

VectorDB are really broken and they just don't work. Excited to try out Hydra!

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

will wait to hear about your experience, my DMs are always open:)

Siddhartha Saxena profil fotoğrafı

Siddhartha Saxena6 ay önce

Yes, going to this a try this weekend!

cole murray profil fotoğrafı

cole murray6 ay önce

man with db tech to sell says other DB tech is bad ignoring the larger search problem

AshutoshShrivastava profil fotoğrafı

AshutoshShrivastava6 ay önce

No 1 failure point is retrieval and most of us on X is arguing mostly about benchmarks. Lol

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

benchmarks won't save you when retrieval grabs the wrong client's contract

Shruti profil fotoğrafı

Shruti6 ay önce

99% of AI companies are building on fundamentally broken retrieval. The 1% who figure this out first will own their vertical.

Tulsi Soni profil fotoğrafı

Tulsi Soni6 ay önce

Actually useful vs the usual "10 AI tools" threads.

Lisa profil fotoğrafı

Lisa6 ay önce

Every company building AI agents needs this. Some of my most gratifying moments as an investor have been hearing how much your customers love the product.

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

hearing things like that from customers is the best part of this whole journey 🙏

Prateek Sharma // Ahead VC profil fotoğrafı

Prateek Sharma // Ahead VC6 ay önce

Love the idea of the librarian to capture the power of hydradb. Humans remember through a web of connections and our memory evolves as context changes. Great to see that finally AI can do the same! Congrats on the launch @contextkingceo and @hydra_db team!

Nishkarsh profil fotoğrafı

Nishkarsh6 ay önce

@hydra_db grateful for your support always @prateeks!

Benzer Videolar

Traditional (SQL) databases rely primarily on keyword-based searches to retrieve information. These searches match the exact words or phrases in your query to the text stored in the database. While effective for many applications, this method has limitations when it comes to understanding context or finding relevant information that doesn’t include the exact keywords. Hybrid search combines the strengths of traditional keyword-based BM25 search with the advanced capabilities of semantic search. To effectively implement a hybrid search, a vector database is essential. Vector databases go beyond just words; they understand the meaning behind the data. They transform data such as text, images, or audio into numerical representations called vectors. These vector embeddings enable the database to find similar items, even if they don't share exact keywords. When you integrate hybrid search with Retrieval-Augmented Generation (RAG) systems, you can achieve higher accuracy in retrieved context and better output in generated responses. Learn more about RAG systems in this video with Victoria Slocum:

Traditional (SQL) databases rely primarily on keyword-based searches to retrieve information. These searches match the exact words or phrases in your query to the text stored in the database. While effective for many applications, this method has limitations when it comes to understanding context or finding relevant information that doesn’t include the exact keywords. Hybrid search combines the strengths of traditional keyword-based BM25 search with the advanced capabilities of semantic search. To effectively implement a hybrid search, a vector database is essential. Vector databases go beyond just words; they understand the meaning behind the data. They transform data such as text, images, or audio into numerical representations called vectors. These vector embeddings enable the database to find similar items, even if they don't share exact keywords. When you integrate hybrid search with Retrieval-Augmented Generation (RAG) systems, you can achieve higher accuracy in retrieved context and better output in generated responses. Learn more about RAG systems in this video with Victoria Slocum:

Femke Plantinga

144,343 görüntüleme • 2 yıl önce

Researchers built a new RAG approach that: - does not need a vector DB. - does not embed data. - involves no chunking. - performs no similarity search. And it hit 98.7% accuracy on a financial benchmark (SOTA). Here's the core problem with RAG that this new approach solves: Traditional RAG chunks documents, embeds them into vectors, and retrieves based on semantic similarity. But similarity ≠ relevance. When you ask "What were the debt trends in 2023?", a vector search returns chunks that look similar. But the actual answer might be buried in some Appendix, referenced on some page, in a section that shares zero semantic overlap with your query. Traditional RAG would likely never find it. PageIndex (open-source) solves this. Instead of chunking and embedding, PageIndex builds a hierarchical tree structure from your documents, like an intelligent table of contents. Then it uses reasoning to traverse that tree. For instance, the model doesn't ask: "What text looks similar to this query?" Instead, it asks: "Based on this document's structure, where would a human expert look for this answer?" That's a fundamentally different approach with: - No arbitrary chunking that breaks context. - No vector DB infrastructure to maintain. - Traceable retrieval to see exactly why it chose a specific section. - The ability to see in-document references ("see Table 5.3") the way a human would. But here's the deeper issue that it solves. Vector search treats every query as independent. But documents have structure and logic, like sections that reference other sections and context that builds across pages. PageIndex respects that structure instead of flattening it into embeddings. Do note that this approach may not make sense in every use case since traditional vector search is still fast, simple, and works well for many applications. But for professional documents that require domain expertise and multi-step reasoning, this tree-based, reasoning-first approach shines. For instance, PageIndex achieved 98.7% accuracy on FinanceBench, significantly outperforming traditional vector-based RAG systems on complex financial document analysis. Everything is fully open-source, so you can see the full implementation in GitHub and try it yourself. I have shared the GitHub repo in the replies!

Researchers built a new RAG approach that: - does not need a vector DB. - does not embed data. - involves no chunking. - performs no similarity search. And it hit 98.7% accuracy on a financial benchmark (SOTA). Here's the core problem with RAG that this new approach solves: Traditional RAG chunks documents, embeds them into vectors, and retrieves based on semantic similarity. But similarity ≠ relevance. When you ask "What were the debt trends in 2023?", a vector search returns chunks that look similar. But the actual answer might be buried in some Appendix, referenced on some page, in a section that shares zero semantic overlap with your query. Traditional RAG would likely never find it. PageIndex (open-source) solves this. Instead of chunking and embedding, PageIndex builds a hierarchical tree structure from your documents, like an intelligent table of contents. Then it uses reasoning to traverse that tree. For instance, the model doesn't ask: "What text looks similar to this query?" Instead, it asks: "Based on this document's structure, where would a human expert look for this answer?" That's a fundamentally different approach with: - No arbitrary chunking that breaks context. - No vector DB infrastructure to maintain. - Traceable retrieval to see exactly why it chose a specific section. - The ability to see in-document references ("see Table 5.3") the way a human would. But here's the deeper issue that it solves. Vector search treats every query as independent. But documents have structure and logic, like sections that reference other sections and context that builds across pages. PageIndex respects that structure instead of flattening it into embeddings. Do note that this approach may not make sense in every use case since traditional vector search is still fast, simple, and works well for many applications. But for professional documents that require domain expertise and multi-step reasoning, this tree-based, reasoning-first approach shines. For instance, PageIndex achieved 98.7% accuracy on FinanceBench, significantly outperforming traditional vector-based RAG systems on complex financial document analysis. Everything is fully open-source, so you can see the full implementation in GitHub and try it yourself. I have shared the GitHub repo in the replies!

Avi Chawla

973,546 görüntüleme • 7 ay önce

Vector databases are a scam. Not technically, they do exactly what they say. Return the most cosine-similar string to your query. The scam is the entire industry pretending that's the same thing as relevance. It isn't. Search "Apple." You get the fruit, the company, the watch, and a recipe blog. Your agent picks one at random and calls it retrieval. Your customer calls it broken. Most AI agents shipping right now are duct-taped on top of this. They demo well because demos are easy. They die in production because production is real. HydraDB's Founder Nish (Nishkarsh) said the quiet part out loud — "vector databases suck, similarity is not relevance" — and the demo signups haven't stopped since. He raised $6.5M because he was the first to name what everyone in the room already knew. If your retrieval layer is a flat embedding index, you're not building infrastructure. You're building a liability with a prettier name. 𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒 (00:00) AI Needs Context (01:30) HydraDB Explained (07:41) Vector Search Breaks (09:32) Messaging That Converts (13:41) Writing the Viral Tweet (16:07) Similarity Not Relevance (20:46) POC to Production Gap (35:35) Raising 6.5 Million Fast (39:33) Founder Lesson on Messaging This is a Composio "Agents at Work" podcast, where I chat with founders building the next leap of AI. Follow for more:)

Vector databases are a scam. Not technically, they do exactly what they say. Return the most cosine-similar string to your query. The scam is the entire industry pretending that's the same thing as relevance. It isn't. Search "Apple." You get the fruit, the company, the watch, and a recipe blog. Your agent picks one at random and calls it retrieval. Your customer calls it broken. Most AI agents shipping right now are duct-taped on top of this. They demo well because demos are easy. They die in production because production is real. HydraDB's Founder Nish (Nishkarsh) said the quiet part out loud — "vector databases suck, similarity is not relevance" — and the demo signups haven't stopped since. He raised $6.5M because he was the first to name what everyone in the room already knew. If your retrieval layer is a flat embedding index, you're not building infrastructure. You're building a liability with a prettier name. 𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒 (00:00) AI Needs Context (01:30) HydraDB Explained (07:41) Vector Search Breaks (09:32) Messaging That Converts (13:41) Writing the Viral Tweet (16:07) Similarity Not Relevance (20:46) POC to Production Gap (35:35) Raising 6.5 Million Fast (39:33) Founder Lesson on Messaging This is a Composio "Agents at Work" podcast, where I chat with founders building the next leap of AI. Follow for more:)

Julia Fedorin

186,958 görüntüleme • 4 ay önce

New short course on Building Applications with Vector Databases, taught by Pinecone’s Tim Tully! At the heart of a vector database is the ability to store a collection of vectors and then query against that, meaning input a new vector and find similar ones. This is useful for many AI applications. In this course, you'll learn how to use vector databases to build: (i) Semantic Search: Create a text search tool that goes beyond keyword matching, and instead focuses on the meaning of content. (ii) RAG (retrieval augmented generation): Enhance your LLM output by incorporating context from sources the model wasn't trained on. (iii) Recommender System: Combine semantic search and RAG to recommend topics, and demonstrate it with a news article recommender. (iv) Hybrid Search: Build an application that finds items using both images and descriptive text -- by combining both sparse and dense vector representations of the data -- using an eCommerce dataset as an example. (v) Image Similarity: Use image vector embeddings to create an app to compare facial features, using a database of public figures to determine the likeness between them. (vi) Anomaly Detection: Build an anomaly detection app that identifies unusual patterns in network communication logs. I hope you’ll enjoy learning how to build all these types of applications! Please sign up here:

New short course on Building Applications with Vector Databases, taught by Pinecone’s Tim Tully! At the heart of a vector database is the ability to store a collection of vectors and then query against that, meaning input a new vector and find similar ones. This is useful for many AI applications. In this course, you'll learn how to use vector databases to build: (i) Semantic Search: Create a text search tool that goes beyond keyword matching, and instead focuses on the meaning of content. (ii) RAG (retrieval augmented generation): Enhance your LLM output by incorporating context from sources the model wasn't trained on. (iii) Recommender System: Combine semantic search and RAG to recommend topics, and demonstrate it with a news article recommender. (iv) Hybrid Search: Build an application that finds items using both images and descriptive text -- by combining both sparse and dense vector representations of the data -- using an eCommerce dataset as an example. (v) Image Similarity: Use image vector embeddings to create an app to compare facial features, using a database of public figures to determine the likeness between them. (vi) Anomaly Detection: Build an anomaly detection app that identifies unusual patterns in network communication logs. I hope you’ll enjoy learning how to build all these types of applications! Please sign up here:

Andrew Ng

137,169 görüntüleme • 2 yıl önce

Traditional data pipelines don't work for RAG applications. There are 3 issues with them: 1. Traditional data engineering solutions are optimized to handle structured data. RAG applications rely primarily on unstructured data. 2. The connector ecosystem to load data from unstructured data sources is very immature. 3. Traditional solutions do not offer any way to transform unstructured data into an optimized vector search index. The goal of a RAG Pipeline is to solve these problems. The number one objective is to create a reliable vector search index using factual knowledge and relevant context. This sounds easy, but it's one of the biggest challenges we face when building RAG applications. At a high level, there are four different stages in the architecture of a RAG pipeline: 1. Ingestion: Here is where the pipeline loads the information from the data source. 2. Extraction: Where the pipeline processes the input data and decides how to retrieve the text contained inside them. 3. Transform: Where the pipeline chunks the data and generates document embeddings. 4. Load: Where the pipeline creates a search index in a vector database and loads the document embeddings. There are different rabbit holes at each one of these stages. Here are three of them: 1. Ingesting data once is simple. The hard part is refreshing the vector database whenever the original data source changes. 2. Extracting the content of a plain text document is simple. The hard part is to extract content from complex documents containing tables, images, or cross-references. 3. A simple continual chunking strategy with an overlap is simple. The hard part is to find the optimal strategy for your specific knowledge base and the way you are planning to query it. In the attached video, I'll show you how you can build an enterprise-grade RAG Pipeline that solves every one of the above problems. I'll use Vectorize. They partnered with me on this post. You can use them to build RAG pipelines optimized for accurate context retrieval. If you have a few documents lying around, set up a free account and give it a try.

Traditional data pipelines don't work for RAG applications. There are 3 issues with them: 1. Traditional data engineering solutions are optimized to handle structured data. RAG applications rely primarily on unstructured data. 2. The connector ecosystem to load data from unstructured data sources is very immature. 3. Traditional solutions do not offer any way to transform unstructured data into an optimized vector search index. The goal of a RAG Pipeline is to solve these problems. The number one objective is to create a reliable vector search index using factual knowledge and relevant context. This sounds easy, but it's one of the biggest challenges we face when building RAG applications. At a high level, there are four different stages in the architecture of a RAG pipeline: 1. Ingestion: Here is where the pipeline loads the information from the data source. 2. Extraction: Where the pipeline processes the input data and decides how to retrieve the text contained inside them. 3. Transform: Where the pipeline chunks the data and generates document embeddings. 4. Load: Where the pipeline creates a search index in a vector database and loads the document embeddings. There are different rabbit holes at each one of these stages. Here are three of them: 1. Ingesting data once is simple. The hard part is refreshing the vector database whenever the original data source changes. 2. Extracting the content of a plain text document is simple. The hard part is to extract content from complex documents containing tables, images, or cross-references. 3. A simple continual chunking strategy with an overlap is simple. The hard part is to find the optimal strategy for your specific knowledge base and the way you are planning to query it. In the attached video, I'll show you how you can build an enterprise-grade RAG Pipeline that solves every one of the above problems. I'll use Vectorize. They partnered with me on this post. You can use them to build RAG pipelines optimized for accurate context retrieval. If you have a few documents lying around, set up a free account and give it a try.

Santiago

40,627 görüntüleme • 1 yıl önce

Engineer runs a Kimi K3 memory layer that costs $11 a month and remembers what a $500,000 vector database keeps losing. No embeddings. Four nodes and one rule about what's allowed to be forgotten. He published the whole schema. His version starts from the opposite idea. Memory is not a pile you search. It's a set of claims that expire unless something keeps paying to keep them. Four nodes. Every memory carries a clock someone has to reset: > WRITER - stores a fact with the reason it mattered, never raw text > DECAY - ages every memory down. Silence is deletion > RENEWER - only re-lifts a memory the model actually used again > GRAVE - holds what died, and why nobody reached for it Three nodes keep memory alive. One keeps the dead ones. Recall isn't storage here. It's rent a fact has to keep earning. That's the entire design. When everything is remembered forever, the useful and the stale retrieve identically. He replayed two months of agent context. 90,000 stored facts. 71,000 never retrieved once. The vector store returned all of them on similarity. Similarity graded closeness. Nobody graded whether the memory was ever right. Everyone else stuffs more into the context window and calls it memory. He built a layer that lets a fact die unless it keeps proving itself. The cost isn't storage. It's finding out how much of what your agent "knows" it has never once used. The article below is the full build - node prompts, the decay curve, the renewal rule. Save it. You'll want it open in the other tab.

Engineer runs a Kimi K3 memory layer that costs $11 a month and remembers what a $500,000 vector database keeps losing. No embeddings. Four nodes and one rule about what's allowed to be forgotten. He published the whole schema. His version starts from the opposite idea. Memory is not a pile you search. It's a set of claims that expire unless something keeps paying to keep them. Four nodes. Every memory carries a clock someone has to reset: > WRITER - stores a fact with the reason it mattered, never raw text > DECAY - ages every memory down. Silence is deletion > RENEWER - only re-lifts a memory the model actually used again > GRAVE - holds what died, and why nobody reached for it Three nodes keep memory alive. One keeps the dead ones. Recall isn't storage here. It's rent a fact has to keep earning. That's the entire design. When everything is remembered forever, the useful and the stale retrieve identically. He replayed two months of agent context. 90,000 stored facts. 71,000 never retrieved once. The vector store returned all of them on similarity. Similarity graded closeness. Nobody graded whether the memory was ever right. Everyone else stuffs more into the context window and calls it memory. He built a layer that lets a fact die unless it keeps proving itself. The cost isn't storage. It's finding out how much of what your agent "knows" it has never once used. The article below is the full build - node prompts, the decay curve, the renewal rule. Save it. You'll want it open in the other tab.

wast3

67,639 görüntüleme • 20 gün önce

This Python script helps you better understand how embeddings work for SEO. Input a sentence + a query and BERT will calculate a content "Similarity Score": Search engines use embedding models to translate your content into numeric values. This is how they're able to mathematically determine whether a page on your site is actually relevant to a query someone is searching for. Once both the content and query are run through embeddings - they can calculate the "cosine similarity" or how similar the two entities are to each one. With this Python script, you can actually visualize how cosine similarity works for yourself. To you use you'll simply: 1. Save the Python script to a text file 2. Use your Terminal to run the script 3. You'll be prompted for both a "Sentence" and "Query" 4. Once entered, the script will calculate a "Similarity Score" between the text and query. This is how relevant your sentence is for the target keyword. By running this, you'll see that search engines are able to come with an calculation of content relevance. You'll need to think about the content on your site the same way - which sections have high scores and which ones have low ones? I've linked the Python script in the comments below. No coding knowledge is required and it walks you step by step on how to implement it.

This Python script helps you better understand how embeddings work for SEO. Input a sentence + a query and BERT will calculate a content "Similarity Score": Search engines use embedding models to translate your content into numeric values. This is how they're able to mathematically determine whether a page on your site is actually relevant to a query someone is searching for. Once both the content and query are run through embeddings - they can calculate the "cosine similarity" or how similar the two entities are to each one. With this Python script, you can actually visualize how cosine similarity works for yourself. To you use you'll simply: 1. Save the Python script to a text file 2. Use your Terminal to run the script 3. You'll be prompted for both a "Sentence" and "Query" 4. Once entered, the script will calculate a "Similarity Score" between the text and query. This is how relevant your sentence is for the target keyword. By running this, you'll see that search engines are able to come with an calculation of content relevance. You'll need to think about the content on your site the same way - which sections have high scores and which ones have low ones? I've linked the Python script in the comments below. No coding knowledge is required and it walks you step by step on how to implement it.

Chris Long

13,185 görüntüleme • 2 yıl önce

Announcing a new Coursera course: Retrieval Augmented Generation (RAG) You'll learn to build high performance, production-ready RAG systems in this hands-on, in-depth course created by and taught by , experienced AI and ML engineer, researcher, and educator. RAG is a critical component today of many LLM-based applications in customer support, internal company Q&A systems, even many of the leading chatbots that use web search to answer your questions. This course teaches you in-depth how to make RAG work well. LLMs can produce generic or outdated responses, especially when asked specialized questions not covered in its training data. RAG is the most widely used technique for addressing this. It brings in data from new data sources, such as internal documents or recent news, to give the LLM the relevant context to private, recent, or specialized information. This lets it generate more grounded and accurate responses. In this course, you’ll learn to design and implement every part of a RAG system, from retrievers to vector databases to generation to evals. You’ll learn about the fundamental principles behind RAG and how to optimize it at both the component and whole-system levels. As AI evolves, RAG is evolving too. New models can handle longer context windows, reason more effectively, and can be parts of complex agentic workflows. One exciting growth area is Agentic RAG, in which an AI agent at runtime (rather than it being hardcoded at development time) autonomously decides what data to retrieve, and when/how to go deeper. Even with this evolution, access to high-quality data at runtime is essential, which is why RAG is a key part of so many applications. You'll learn via hands-on experiences to: - Build a RAG system with retrieval and prompt augmentation - Compare retrieval methods like BM25, semantic search, and Reciprocal Rank Fusion - Chunk, index, and retrieve documents using a Weaviate vector database and a news dataset - Develop a chatbot, using open-source LLMs hosted by Together AI, for a fictional store that answers product and FAQ questions - Use evals to drive improving reliability, and incorporate multi-modal data RAG is an important foundational technique. Become good at it through this course! Please sign up here:

Announcing a new Coursera course: Retrieval Augmented Generation (RAG) You'll learn to build high performance, production-ready RAG systems in this hands-on, in-depth course created by and taught by , experienced AI and ML engineer, researcher, and educator. RAG is a critical component today of many LLM-based applications in customer support, internal company Q&A systems, even many of the leading chatbots that use web search to answer your questions. This course teaches you in-depth how to make RAG work well. LLMs can produce generic or outdated responses, especially when asked specialized questions not covered in its training data. RAG is the most widely used technique for addressing this. It brings in data from new data sources, such as internal documents or recent news, to give the LLM the relevant context to private, recent, or specialized information. This lets it generate more grounded and accurate responses. In this course, you’ll learn to design and implement every part of a RAG system, from retrievers to vector databases to generation to evals. You’ll learn about the fundamental principles behind RAG and how to optimize it at both the component and whole-system levels. As AI evolves, RAG is evolving too. New models can handle longer context windows, reason more effectively, and can be parts of complex agentic workflows. One exciting growth area is Agentic RAG, in which an AI agent at runtime (rather than it being hardcoded at development time) autonomously decides what data to retrieve, and when/how to go deeper. Even with this evolution, access to high-quality data at runtime is essential, which is why RAG is a key part of so many applications. You'll learn via hands-on experiences to: - Build a RAG system with retrieval and prompt augmentation - Compare retrieval methods like BM25, semantic search, and Reciprocal Rank Fusion - Chunk, index, and retrieve documents using a Weaviate vector database and a news dataset - Develop a chatbot, using open-source LLMs hosted by Together AI, for a fictional store that answers product and FAQ questions - Use evals to drive improving reliability, and incorporate multi-modal data RAG is an important foundational technique. Become good at it through this course! Please sign up here:

Andrew Ng

124,656 görüntüleme • 1 yıl önce

Is AI being designed to fail? Everyone talks about reasoning. But when given a task, the AI isn't reasoning the way you might expect. It looks at your input, finds the closest match it's seen before, and predicts the most likely next action. That process is called vector similarity search. It's genuinely powerful. It's also not the same thing as understanding what you actually meant. Think of a plumber who hears the word "leak" and starts pulling up floorboards before you've finished the sentence. He's not being careless. He's pattern-matching - that's exactly how he was trained. Your AI agent is doing the same thing. Context is the one thing that gets deprioritized when teams are racing to ship. But without it, you don't have an intelligent agent. You have a very fast guesser. Similarity ≠ relevance. How? Find out with the link in the comments ⬇️

Is AI being designed to fail? Everyone talks about reasoning. But when given a task, the AI isn't reasoning the way you might expect. It looks at your input, finds the closest match it's seen before, and predicts the most likely next action. That process is called vector similarity search. It's genuinely powerful. It's also not the same thing as understanding what you actually meant. Think of a plumber who hears the word "leak" and starts pulling up floorboards before you've finished the sentence. He's not being careless. He's pattern-matching - that's exactly how he was trained. Your AI agent is doing the same thing. Context is the one thing that gets deprioritized when teams are racing to ship. But without it, you don't have an intelligent agent. You have a very fast guesser. Similarity ≠ relevance. How? Find out with the link in the comments ⬇️

Nishkarsh

823,177 görüntüleme • 5 ay önce

Verba is an open source Retrieval Augmented Generation (RAG) application that performs RAG on your own data. To showcase its capabilities, we've customized it as an Airbnb chatbot using Airbnb’s customer documentation. How it works: • Ask any questions, related to your booking, policies, or anything related to your Airbnb experience. • Get relevant, human-like responses: Verba provides natural and informative answers. • Access original sources: One of the standout features of RAG is its ability to directly indicate the sources it used to generate each response. Under the hood, Verba uses a RAG pipeline to deliver these exceptional results. Your query is transformed into a numerical representation (vector) and be used to search through our vector database for the most similar context using Hybrid Search. The most relevant context is then combined with your original question and fed into a powerful large language model (LLM). The LLM will then use all of that information to generate a conversational response. Et voilà! 💫 Try Verba: Verba on GitHub: Learn more in our video:

Verba is an open source Retrieval Augmented Generation (RAG) application that performs RAG on your own data. To showcase its capabilities, we've customized it as an Airbnb chatbot using Airbnb’s customer documentation. How it works: • Ask any questions, related to your booking, policies, or anything related to your Airbnb experience. • Get relevant, human-like responses: Verba provides natural and informative answers. • Access original sources: One of the standout features of RAG is its ability to directly indicate the sources it used to generate each response. Under the hood, Verba uses a RAG pipeline to deliver these exceptional results. Your query is transformed into a numerical representation (vector) and be used to search through our vector database for the most similar context using Hybrid Search. The most relevant context is then combined with your original question and fed into a powerful large language model (LLM). The LLM will then use all of that information to generate a conversational response. Et voilà! 💫 Try Verba: Verba on GitHub: Learn more in our video:

Femke Plantinga

120,565 görüntüleme • 2 yıl önce

If you don't know what a RAG system is, you're already behind. Plenty of people are still typing questions into ChatGPT and pasting answers back out. That's the beginner level. You're renting a smart stranger who forgets you the second you close the tab and makes things up, because it has never seen your actual work. A RAG system is a completely different level of using AI. You build an AI plugged into your own world. Your files, company's docs, notes. It's not pulling from the internet, it only knows what you fed it. Ask it anything and it digs through your data, finds the answer, and shows you which document it came from. That's the gap. One person is asking a chatbot for generic answers. The other has built a private expert that knows their business cold and answers with receipts. And that gap is worth real money: > At work, the person who can wire a company's knowledge into an AI that actually answers correctly is rare and paid like it. > As a business, companies pay serious money for a system that reads their internal docs and answers staff and customers accurately. You build it once, you sell it. > For yourself, you get a second brain that actually remembers. Same skill behind all three, and it's still rare enough to be worth learning. Full breakdown in the article below.

If you don't know what a RAG system is, you're already behind. Plenty of people are still typing questions into ChatGPT and pasting answers back out. That's the beginner level. You're renting a smart stranger who forgets you the second you close the tab and makes things up, because it has never seen your actual work. A RAG system is a completely different level of using AI. You build an AI plugged into your own world. Your files, company's docs, notes. It's not pulling from the internet, it only knows what you fed it. Ask it anything and it digs through your data, finds the answer, and shows you which document it came from. That's the gap. One person is asking a chatbot for generic answers. The other has built a private expert that knows their business cold and answers with receipts. And that gap is worth real money: > At work, the person who can wire a company's knowledge into an AI that actually answers correctly is rare and paid like it. > As a business, companies pay serious money for a system that reads their internal docs and answers staff and customers accurately. You build it once, you sell it. > For yourself, you get a second brain that actually remembers. Same skill behind all three, and it's still rare enough to be worth learning. Full breakdown in the article below.

Yarchi

32,840 görüntüleme • 2 ay önce

The United States Officially Activates AI Warfare Why was a US-based AI company paying Nigerians for personal data and information from their surroundings? That is the bigger question behind Kled’s decision to ban Nigeria from its platform over alleged fraud-related activities. It is not a question about whether or not some users broke the rules. Why was a foreign AI company operating a business model that offered financial incentives in exchange for the personal data and environmental information of Nigerians in the first place? At a time when U.S. tech companies are becoming increasingly tied to the country’s military and intelligence outfits – if there ever even was a time when they weren’t – Africans must treat data extraction as a serious sovereignty issue. AI platforms cannot be considered neutral tools when they are built inside systems that serve foreign political, economic, and military interests. Data is power. In the wrong hands, it becomes weaponizable intelligence. Biggest Mack reports for the Spearhead.

The United States Officially Activates AI Warfare Why was a US-based AI company paying Nigerians for personal data and information from their surroundings? That is the bigger question behind Kled’s decision to ban Nigeria from its platform over alleged fraud-related activities. It is not a question about whether or not some users broke the rules. Why was a foreign AI company operating a business model that offered financial incentives in exchange for the personal data and environmental information of Nigerians in the first place? At a time when U.S. tech companies are becoming increasingly tied to the country’s military and intelligence outfits – if there ever even was a time when they weren’t – Africans must treat data extraction as a serious sovereignty issue. AI platforms cannot be considered neutral tools when they are built inside systems that serve foreign political, economic, and military interests. Data is power. In the wrong hands, it becomes weaponizable intelligence. Biggest Mack reports for the Spearhead.

The Spearhead

90,491 görüntüleme • 4 ay önce

For 40 years the file browser hasn’t changed. Today, we’re launching with $8 million in seed funding to rebuild the file browser into something more intelligent, searchable, and delightful. The world is in the middle of a data explosion. We’re generating and using more files than ever, but the apps we’re using to manage our files don’t even understand them. It’s time for file browsers to become useful. When you search for “dog”, it should show you content with dogs in it, not just files with “dog” in the name! When you want to edit, convert, summarize, or organize a file, your browser should do that, too. Your files tell the story of your life, but when you need a specific one, you usually can’t even find it anymore. Why can’t your file browser find it for you, or cross-reference it when you have a question? Prompting can give an LLM a million tokens of context. With Poly, you can give it the next trillion. As long as we can afford it, all new users receive 100GB of free cloud storage. We can’t wait for you to try it out!

For 40 years the file browser hasn’t changed. Today, we’re launching with $8 million in seed funding to rebuild the file browser into something more intelligent, searchable, and delightful. The world is in the middle of a data explosion. We’re generating and using more files than ever, but the apps we’re using to manage our files don’t even understand them. It’s time for file browsers to become useful. When you search for “dog”, it should show you content with dogs in it, not just files with “dog” in the name! When you want to edit, convert, summarize, or organize a file, your browser should do that, too. Your files tell the story of your life, but when you need a specific one, you usually can’t even find it anymore. Why can’t your file browser find it for you, or cross-reference it when you have a question? Prompting can give an LLM a million tokens of context. With Poly, you can give it the next trillion. As long as we can afford it, all new users receive 100GB of free cloud storage. We can’t wait for you to try it out!

Abhay Agarwal

1,777,939 görüntüleme • 10 ay önce

Anthropic just got outplayed again. Devs built the multiplayer assistant Anthropic couldn't, and open-sourced it. Claude Cowork is a solo desktop agent. You point it at a folder, give it a task, and it works through your local files on your own machine. The moment a teammate enters the picture, it has nothing to offer. Most real work does not happen alone. A teammate asks for a status update on something you own. The context they need is scattered across your meetings, your notes, and decisions made last week. Typing all of that out takes time you do not have. This is the gap Claude Cowork was never designed to cross. Rowboat Spaces is built on a different model entirely. Each person brings their own assistant into a shared channel. Your assistant is your second brain. It knows your meetings, your notes, and your open decisions. That personal context stays yours. When a teammate asks a question in the channel, you ask your assistant to brief them. It pulls from everything you know and delivers the answer on your behalf, attributed to you. Your teammate's assistant does the same, from their own context. Teams can draft specs, track decisions, and update shared files from plain conversation. Each assistant reads the full channel history, cross references it against what exists, and flags what is missing. The whole thing is open-source, and each assistant acts as the person it belongs to, not as a shared bot pulling from a common pool. The video below shows this in action. I joined a shared space and asked my team member for a status update. My team member asked their second brain to answer. A spec got built from that conversation, versioned, with every change tracked back to the message that triggered it. Rowboat GitHub: (don't forget to star 🌟) My co-founder also wrote a great article on building your second brain with Rowboat, and I highly recommend reading it as well. The article is quoted below.

Anthropic just got outplayed again. Devs built the multiplayer assistant Anthropic couldn't, and open-sourced it. Claude Cowork is a solo desktop agent. You point it at a folder, give it a task, and it works through your local files on your own machine. The moment a teammate enters the picture, it has nothing to offer. Most real work does not happen alone. A teammate asks for a status update on something you own. The context they need is scattered across your meetings, your notes, and decisions made last week. Typing all of that out takes time you do not have. This is the gap Claude Cowork was never designed to cross. Rowboat Spaces is built on a different model entirely. Each person brings their own assistant into a shared channel. Your assistant is your second brain. It knows your meetings, your notes, and your open decisions. That personal context stays yours. When a teammate asks a question in the channel, you ask your assistant to brief them. It pulls from everything you know and delivers the answer on your behalf, attributed to you. Your teammate's assistant does the same, from their own context. Teams can draft specs, track decisions, and update shared files from plain conversation. Each assistant reads the full channel history, cross references it against what exists, and flags what is missing. The whole thing is open-source, and each assistant acts as the person it belongs to, not as a shared bot pulling from a common pool. The video below shows this in action. I joined a shared space and asked my team member for a status update. My team member asked their second brain to answer. A spec got built from that conversation, versioned, with every change tracked back to the message that triggered it. Rowboat GitHub: (don't forget to star 🌟) My co-founder also wrote a great article on building your second brain with Rowboat, and I highly recommend reading it as well. The article is quoted below.

Akshay 🚀

105,135 görüntüleme • 1 gün önce

For 60 years every computer ever built did the same thing. Stored information and retrieved it on demand. Jensen Huang just explained why that era is over and what replaces it. His framing was the clearest I have ever heard. Think about everything a computer has ever done for you. You wrote a document, you saved it to a file. You took a photo, it saved to a file. You recorded music, it saved to a file. When you wanted it back, you retrieved it from a disc. That is it. That is 60 years of computing. Store and retrieve. He pointed out something hiding in plain sight. We call them data centers. Not computer centers. Because we were not really computing anything meaningful. We were storing data that you retrieved based on what you tapped on your phone. Then he explained what changed. Every time you give AI a prompt today, the response is produced originally in real time. It is not retrieved from storage. It is generated fresh based on your specific context, your specific question, your specific moment. What you see is completely different from what anyone else sees because it was made for you. Jensen said every pixel you see, every word you read, every video you watch in the future will be originally generated. Not retrieved. 60 years of computing was about building better storage and faster retrieval. The entire paradigm flipped overnight. He said this simply: we went from a retrieval industry to a generation industry. And the machines that generate intelligence are what Nvidia builds. The buildings used to be called data centers because they stored data. Nobody has renamed them yet. But the job description changed completely.

For 60 years every computer ever built did the same thing. Stored information and retrieved it on demand. Jensen Huang just explained why that era is over and what replaces it. His framing was the clearest I have ever heard. Think about everything a computer has ever done for you. You wrote a document, you saved it to a file. You took a photo, it saved to a file. You recorded music, it saved to a file. When you wanted it back, you retrieved it from a disc. That is it. That is 60 years of computing. Store and retrieve. He pointed out something hiding in plain sight. We call them data centers. Not computer centers. Because we were not really computing anything meaningful. We were storing data that you retrieved based on what you tapped on your phone. Then he explained what changed. Every time you give AI a prompt today, the response is produced originally in real time. It is not retrieved from storage. It is generated fresh based on your specific context, your specific question, your specific moment. What you see is completely different from what anyone else sees because it was made for you. Jensen said every pixel you see, every word you read, every video you watch in the future will be originally generated. Not retrieved. 60 years of computing was about building better storage and faster retrieval. The entire paradigm flipped overnight. He said this simply: we went from a retrieval industry to a generation industry. And the machines that generate intelligence are what Nvidia builds. The buildings used to be called data centers because they stored data. Nobody has renamed them yet. But the job description changed completely.

Ihtesham Ali

30,235 görüntüleme • 3 ay önce

How can you solve complex tasks using a Large Language Model? Here is a 2-minute introduction to everything you need to know to 10x the quality of your results. Let's talk about three techniques, in order of complexity, starting with the easiest one: • In-Context Learning • Indexing + In-Context Learning • Fine-tuning In-Context Learning The team that trained GPT-3 found something they couldn't explain: You can condition a model using examples of how you want it to behave. I included an example prompt in the attached video. You can "teach" the model how you want it to interpret questions, select the correct answers, and format the results by giving a few examples. You can also give specific knowledge to the model that will be helpful when formulating answers. We call this approach "grounding the model." There's another example in the video. Indexing + In-Context Learning Unfortunately, there is a limit to how much data you can include in a prompt. We call this the "context size." One version of GPT-4 supports a context of approximately 6,000 words, while the other supports 25,000 words. Although this sounds like a lot, many applications need more than that. Imagine you wrote a book and want to build an application to answer any questions about your story. What happens if your book is longer than the context? That's where Indexing comes in. Using a model, you can turn every book passage into an embedding. These are vectors, numbers that "encode" the passage's text. You can then store these embeddings in a particular database that supports fast retrieval of these vectors. You can then turn any question into an embedding and search the database for the list of passages that are similar to that query. Instead of using the entire book to ask the model, you can now use the relevant passages as in-context information, effectively working around the context size limitation. Fine-tuning Fine-tuning can give you an extra boost to get reliable outputs from your LLM. It is, however, the most complex approach on the list. There are different approaches to fine-tuning a model with your data. A popular technique is to process your data with your LLM and use the outputs to train a new classifier that solves your specific task. Notice that here you aren't modifying the LLM. Instead, you are chaining it with your trained classifier. Another approach is to modify the parameters of the LLM using your data. Think of this as "rewiring" the model in a way that solves your particular task. The results and costs will vary depending on how many layers you want to fine-tune from the original model. Many companies think that fine-tuning is the solution to their problems. In my experience, many will benefit from exploring the other two approaches. I love explaining Machine Learning and Artificial Intelligence ideas. If you enjoy in-depth content like this, follow me Santiago so you don't miss what comes next.

How can you solve complex tasks using a Large Language Model? Here is a 2-minute introduction to everything you need to know to 10x the quality of your results. Let's talk about three techniques, in order of complexity, starting with the easiest one: • In-Context Learning • Indexing + In-Context Learning • Fine-tuning In-Context Learning The team that trained GPT-3 found something they couldn't explain: You can condition a model using examples of how you want it to behave. I included an example prompt in the attached video. You can "teach" the model how you want it to interpret questions, select the correct answers, and format the results by giving a few examples. You can also give specific knowledge to the model that will be helpful when formulating answers. We call this approach "grounding the model." There's another example in the video. Indexing + In-Context Learning Unfortunately, there is a limit to how much data you can include in a prompt. We call this the "context size." One version of GPT-4 supports a context of approximately 6,000 words, while the other supports 25,000 words. Although this sounds like a lot, many applications need more than that. Imagine you wrote a book and want to build an application to answer any questions about your story. What happens if your book is longer than the context? That's where Indexing comes in. Using a model, you can turn every book passage into an embedding. These are vectors, numbers that "encode" the passage's text. You can then store these embeddings in a particular database that supports fast retrieval of these vectors. You can then turn any question into an embedding and search the database for the list of passages that are similar to that query. Instead of using the entire book to ask the model, you can now use the relevant passages as in-context information, effectively working around the context size limitation. Fine-tuning Fine-tuning can give you an extra boost to get reliable outputs from your LLM. It is, however, the most complex approach on the list. There are different approaches to fine-tuning a model with your data. A popular technique is to process your data with your LLM and use the outputs to train a new classifier that solves your specific task. Notice that here you aren't modifying the LLM. Instead, you are chaining it with your trained classifier. Another approach is to modify the parameters of the LLM using your data. Think of this as "rewiring" the model in a way that solves your particular task. The results and costs will vary depending on how many layers you want to fine-tune from the original model. Many companies think that fine-tuning is the solution to their problems. In my experience, many will benefit from exploring the other two approaches. I love explaining Machine Learning and Artificial Intelligence ideas. If you enjoy in-depth content like this, follow me Santiago so you don't miss what comes next.

Santiago

384,510 görüntüleme • 3 yıl önce

context engineering vs graph engineering. every few months the list gets a new word and everyone treats it as a replacement for the last one. these two are not on the same list. one decides what the model sees this turn, the other decides what exists at all. the cleanest way to tell them apart is to ask what a single unit of work looks like. > context engineering is the window the window opens empty, every single time. you assemble what goes in it. the prompt, the docs, the history, the tool results. the assembling is the work. the window only grows. it never shrinks on its own, so eventually something gets dropped. usually from the middle. usually without telling you. then the turn ends and the window is thrown away. not archived, thrown away. the next turn opens empty again and you re-explain what you already explained. good context engineering is knowing what to leave out, not what to pack in. the unit of work is one window. > graph engineering is the structure the same material arrives from the same sources. instead of packing it into a window, you pull entities out of it, resolve the duplicates into one node, and write typed edges between them. nothing here is stored as text you hope to find again. it is stored as a thing with a name and its connections to other things. when the turn ends, the graph is still there. the next turn does not start from zero. it starts by querying what already exists, and the query walks edges instead of guessing at similarity. good graph engineering is deciding what counts as the same thing twice. the unit of work is one relationship. > they are not alternatives the graph is what refills the window. context engineering decides what fits. graph engineering decides what there is to choose from. remove the graph and every session starts blind. remove the context work and the best structure in the world arrives as an unreadable dump. that also tells you which one broke. the answer drifted from what you actually said, or forgot something from this same session. that is the window. the answer is coherent but invents a connection that does not exist, or cannot join two facts it has clearly seen. that is the structure. people debug the prompt because the prompt is the easiest thing to edit. it keeps taking the blame for failures that live a layer down. save this - then read the full breakdown below

context engineering vs graph engineering. every few months the list gets a new word and everyone treats it as a replacement for the last one. these two are not on the same list. one decides what the model sees this turn, the other decides what exists at all. the cleanest way to tell them apart is to ask what a single unit of work looks like. > context engineering is the window the window opens empty, every single time. you assemble what goes in it. the prompt, the docs, the history, the tool results. the assembling is the work. the window only grows. it never shrinks on its own, so eventually something gets dropped. usually from the middle. usually without telling you. then the turn ends and the window is thrown away. not archived, thrown away. the next turn opens empty again and you re-explain what you already explained. good context engineering is knowing what to leave out, not what to pack in. the unit of work is one window. > graph engineering is the structure the same material arrives from the same sources. instead of packing it into a window, you pull entities out of it, resolve the duplicates into one node, and write typed edges between them. nothing here is stored as text you hope to find again. it is stored as a thing with a name and its connections to other things. when the turn ends, the graph is still there. the next turn does not start from zero. it starts by querying what already exists, and the query walks edges instead of guessing at similarity. good graph engineering is deciding what counts as the same thing twice. the unit of work is one relationship. > they are not alternatives the graph is what refills the window. context engineering decides what fits. graph engineering decides what there is to choose from. remove the graph and every session starts blind. remove the context work and the best structure in the world arrives as an unreadable dump. that also tells you which one broke. the answer drifted from what you actually said, or forgot something from this same session. that is the window. the answer is coherent but invents a connection that does not exist, or cannot join two facts it has clearly seen. that is the structure. people debug the prompt because the prompt is the easiest thing to edit. it keeps taking the blame for failures that live a layer down. save this - then read the full breakdown below

Hanako

19,160 görüntüleme • 1 ay önce

how to use Google's NEW open source Design.md + AI Skills to make your startup look like a $100 million company in 1 hour: 1. Design.md is an open source file from Google that captures the soul of a design. Typography, colors, spacing, all in one markdown file. You attach it to your prompt and your agent builds beautiful things every time. 2. Think of it this way. The HTML is the finished dish. The design.md is the recipe. The skills are the ingredients. Put them together and everything you build looks consistent and professional. 3. Don't create a design system from scratch. Find a brand you love. Linear, Stripe, Vercel, whatever resonates. Study it. Use ChatGPT or Claude to help you extract the design language into your own design.md file. 4. Build skills on top of your design.md. A landing page skill. A mobile app skill. A motion design skill. A slide deck skill. Each one references the same design.md so everything looks like it came from the same designer. 5. The biggest mistake people make: they nail one screen and then everything else looks generic. Design.md solves this. One file keeps every page, every format, every medium consistent. 6. Use it across everything. Your landing page. Your app. Your pitch deck. Your promo videos. Same DNA. Same taste. Same system. That's what separates a startup that looks real from one that looks vibe-coded. 7. Build a second brain for design inspiration. When you see something beautiful in the real world or online, capture it. Save it. When you're building something new, reference it. Taste is developed, not downloaded. 8. It's obvious but the difference between a product people trust and a product people bounce from is how it looks and feels. Design.md gives you that edge. you can watch below shoutout to Meng To for coming on The Startup Ideas Podcast (SIP) 🧃 and walking through his full workflow. if you want to use AI to actually build gorgeous designs, you'll want to use see this. watch

GREG ISENBERG

511,448 görüntüleme • 4 ay önce

MCP is an absolute game-changer. (Together with DeepSeek, MCP is probably the hottest thing in AI over the last 6 months.) I use Cursor to write code 90% of the time. I built an MCP server to connect the Cursor agent to GroundX, an open-source RAG system, and I'm not going back. This is officially insane! Here is what I did, step by step: First, a little bit of context. I maintain an end-to-end Machine Learning System with several pipelines to process data, train, evaluate, register, deploy, and monitor a model. I've written a lot of documentation explaining how the system works and how to modify and maintain it. There's also the documentation of the few libraries I used to build the system. I'm a massive fan of GroundX, an open-source enterprise-grade RAG system you can run on your servers or deploy to any cloud provider. I've been working with them for a long time. GroundX offers two services. First, the "ingest" service uses a custom, pretrained vision model to ingest and understand your data. I used this to process all the documentation I have for my code. Markdown files, source code, HTML files, and even PDF documents. Everything I've written related to my project went into GroundX. Their second service is "search," which combines text and vector search with a fine-tuned re-ranker model to retrieve information from the data. I needed to connect Cursor with this service, and that's where MCP came in. I built an MCP server with two tools: 1. The first tool would go to GroundX and retrieve the available topics. Splitting the data into topics (or "buckets," as GroundX calls them) allows me to use the same setup to serve documentation from different topics. 2. The second tool would search GroundX under a specific topic for the context related to the supplied query. The magic happens after connecting the MCP server with Cursor. Now, I can ask any questions related to my project, and Cursor's AI agent retrieves the list of available topics from the RAG system and then searches it to provide relevant context to the model. I went from getting mediocre, sometimes wrong answers to 100% truthful, complete answers. Here is the crazy part:

MCP is an absolute game-changer. (Together with DeepSeek, MCP is probably the hottest thing in AI over the last 6 months.) I use Cursor to write code 90% of the time. I built an MCP server to connect the Cursor agent to GroundX, an open-source RAG system, and I'm not going back. This is officially insane! Here is what I did, step by step: First, a little bit of context. I maintain an end-to-end Machine Learning System with several pipelines to process data, train, evaluate, register, deploy, and monitor a model. I've written a lot of documentation explaining how the system works and how to modify and maintain it. There's also the documentation of the few libraries I used to build the system. I'm a massive fan of GroundX, an open-source enterprise-grade RAG system you can run on your servers or deploy to any cloud provider. I've been working with them for a long time. GroundX offers two services. First, the "ingest" service uses a custom, pretrained vision model to ingest and understand your data. I used this to process all the documentation I have for my code. Markdown files, source code, HTML files, and even PDF documents. Everything I've written related to my project went into GroundX. Their second service is "search," which combines text and vector search with a fine-tuned re-ranker model to retrieve information from the data. I needed to connect Cursor with this service, and that's where MCP came in. I built an MCP server with two tools: 1. The first tool would go to GroundX and retrieve the available topics. Splitting the data into topics (or "buckets," as GroundX calls them) allows me to use the same setup to serve documentation from different topics. 2. The second tool would search GroundX under a specific topic for the context related to the supplied query. The magic happens after connecting the MCP server with Cursor. Now, I can ask any questions related to my project, and Cursor's AI agent retrieves the list of available topics from the RAG system and then searches it to provide relevant context to the model. I went from getting mediocre, sometimes wrong answers to 100% truthful, complete answers. Here is the crazy part:

Santiago

255,723 görüntüleme • 1 yıl önce