正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

I built ChatGPT's memory system into Apple's LLM

Steven Zhang

11,562 subscribers

120,124 次观看 • 10 个月前 •via X (Twitter)

教育科学技术

Anya Rossi• Live Now

Private livecam show

24 条评论

Sanchit monga 的头像

Sanchit monga10 个月前

Checkout @RunAnywhereAI we support Apple foundation models and other inference engines, and also bringing support for embeddings

hinzi 的头像

hinzi10 个月前

Hey man I built a similar app but for llms via openrouter and with council mode. Would love to connect!

Sahil 的头像

Sahil10 个月前

Very cool Steve. Not sure what you mean by apple’s llm though ? Is this an open source model or something Apple is already shipping in its devices ?

Steven Zhang 的头像

Steven Zhang10 个月前

it’s a low parameter model that’s available in iOS 26

Stevie Graham (new account) 的头像

Stevie Graham (new account)10 个月前

Nice work man. I think @therealdaneel wants an orange McLaren too!

Steven Zhang 的头像

Steven Zhang10 个月前

@therealdaneel let’s see who can get one first

Ego Sum Banditos 的头像

Ego Sum Banditos10 个月前

Great job !

Jeff 的头像

Jeff10 个月前

Very cool

Prasanna 的头像

Prasanna10 个月前

@LocallyAIApp should get this

Daniel 的头像

Daniel10 个月前

Ohh that's cool How do I try this?

Josué Gdm 的头像

Josué Gdm10 个月前

wait this is fire

toma 的头像

toma10 个月前

have you tried any of the memory benchmarks?

Christopher Karani 的头像

Christopher Karani10 个月前

would love to try it out

Justin Liao 的头像

Justin Liao10 个月前

Wow niche, I’d love to try

lenny 🥷 的头像

lenny 🥷9 个月前

cool demo - digging the lightweight summaries for on-device personalization. keeps ai snappy without rag. just followed - let's connect! 🙃🤝

enzo enrico 的头像

enzo enrico10 个月前

wait, how did you manage to make it so the model remembers stuff past its (very low) context window?

Kaya Jones 的头像

Kaya Jones10 个月前

share the github

LMD 的头像

LMD10 个月前

You can just do things

Godfrey.Builds 的头像

Godfrey.Builds10 个月前

what would be some usecases for the model? Is it just a free llm app?

Sambit 的头像

Sambit10 个月前

@rudrank you should have look at this!

Ron Adin 的头像

Ron Adin10 个月前

this is good! pls see dm

mscode07 的头像

mscode0710 个月前

Sounds interesting

Yan-David (Yanda) Erlich 的头像

Yan-David (Yanda) Erlich10 个月前

👀😍

Mher 的头像

Mher10 个月前

Is this using an on-device LLM provided by Apple?

相关视频

I turned my iPad into Tom Riddle’s diary with Apple’s new Foundation Models, an LLM built into iOS. It’s lightning fast!

I turned my iPad into Tom Riddle’s diary with Apple’s new Foundation Models, an LLM built into iOS. It’s lightning fast!

Ethan Saadia @ SIGGRAPH

330,886 次观看 • 1 年前

Holy shit... One month before Andrej Karpathy dropped the “LLM Wiki” idea… someone had already built it. Not theory. Not hype. A working system. LLMs today restart from zero every time. No memory. No evolution. No learning. That’s broken. Most people use CLAUDE.md / AGENTS.md → Manual updates → Easy to forget → Knowledge gets lost That’s not intelligence. That’s a notepad. So I built dynamic memory for LLMs: • Learns automatically • Stores without manual effort • Weakens irrelevant data over time • Strengthens what actually matters Before the “LLM Wiki” idea went viral, this already had: • Brain-like decay/boost memory • Smart deduplication (no noise) • Experience tracking per tech • Cross-referenced knowledge graph • Loss-proof backups • Auto-grouped tool sessions Then I added: • memory_audit (clean + optimize memory) • Memory injection directly into tools Now this isn’t just a concept. It’s a live system that turns LLMs from stateless tools → evolving intelligence. Using it daily: → ~5K tokens → Zero manual work → Memory that actually compounds Link in comments. RT & Bookmark — this will be standard soon.

Holy shit... One month before Andrej Karpathy dropped the “LLM Wiki” idea… someone had already built it. Not theory. Not hype. A working system. LLMs today restart from zero every time. No memory. No evolution. No learning. That’s broken. Most people use CLAUDE.md / AGENTS.md → Manual updates → Easy to forget → Knowledge gets lost That’s not intelligence. That’s a notepad. So I built dynamic memory for LLMs: • Learns automatically • Stores without manual effort • Weakens irrelevant data over time • Strengthens what actually matters Before the “LLM Wiki” idea went viral, this already had: • Brain-like decay/boost memory • Smart deduplication (no noise) • Experience tracking per tech • Cross-referenced knowledge graph • Loss-proof backups • Auto-grouped tool sessions Then I added: • memory_audit (clean + optimize memory) • Memory injection directly into tools Now this isn’t just a concept. It’s a live system that turns LLMs from stateless tools → evolving intelligence. Using it daily: → ~5K tokens → Zero manual work → Memory that actually compounds Link in comments. RT & Bookmark — this will be standard soon.

Nainsi Dwivedi

15,814 次观看 • 6 个月前

How to hack apple's emoji system:

Sensitive content

How to hack apple's emoji system:

AlphaFox

4,391,664 次观看 • 1 年前

He is 17 sold an LLM system to Anthropic for $2.2M - and at Stanford shared how to build and sell one from scratch: 00:20 - how to build an LLM system that Anthropic will buy for millions 08:34 - one prompt in Opus 5 and the LLM system is ready to sell in 60 minutes 23:47 - how to sell an LLM to Anthropic at 17 with no experience and reach $2.2M a year after watching I built my first LLM system in 60 minutes - and sold it to Anthropic for $120k. save & watch - the article below is one file that builds an LLM system and sells it to Anthropic for $2.2M.

He is 17 sold an LLM system to Anthropic for $2.2M - and at Stanford shared how to build and sell one from scratch: 00:20 - how to build an LLM system that Anthropic will buy for millions 08:34 - one prompt in Opus 5 and the LLM system is ready to sell in 60 minutes 23:47 - how to sell an LLM to Anthropic at 17 with no experience and reach $2.2M a year after watching I built my first LLM system in 60 minutes - and sold it to Anthropic for $120k. save & watch - the article below is one file that builds an LLM system and sells it to Anthropic for $2.2M.

Noisy

25,571 次观看 • 1 个月前

I built an LLM from scratch. Here's everything I learned along the way.

I built an LLM from scratch. Here's everything I learned along the way.

CJ (Coding Garden)

258,519 次观看 • 3 个月前

I needed a quick tool to convert any sort of GitHub links into a format an LLM could read. So I built it. GitHub → LLM It's completely free and works with projects, folders, or files 🪄 Deployed on Replit ⠕

I needed a quick tool to convert any sort of GitHub links into a format an LLM could read. So I built it. GitHub → LLM It's completely free and works with projects, folders, or files 🪄 Deployed on Replit ⠕

Pietro Schirano

120,827 次观看 • 2 年前

He is a 1st year AI engineering student at Stanford - and already launched an automated system making $2.1M a year: 00:20 - how Anthropic engineers make millions with LLM and Transformers 38:34 - In-Context Learning makes AI smarter for free 59:47 - 1st year student made $2.1M using LLM from scratch after watching I gave LLM the idea I had been holding in my head for 3 years - it built it in one evening. save & watch - the article below is a step by step guide to build your automated LLM system.

He is a 1st year AI engineering student at Stanford - and already launched an automated system making $2.1M a year: 00:20 - how Anthropic engineers make millions with LLM and Transformers 38:34 - In-Context Learning makes AI smarter for free 59:47 - 1st year student made $2.1M using LLM from scratch after watching I gave LLM the idea I had been holding in my head for 3 years - it built it in one evening. save & watch - the article below is a step by step guide to build your automated LLM system.

Noisy

43,713 次观看 • 2 个月前

I built the LLM interface I wished existed. interface0 is for power users: cross-provider memory & chat, multi-model synthesis, sharing & team features, granular context engineering, template prompts, forking, prompt enhancing, & more Here's a demo—see 🧵 for the rest

I built the LLM interface I wished existed. interface0 is for power users: cross-provider memory & chat, multi-model synthesis, sharing & team features, granular context engineering, template prompts, forking, prompt enhancing, & more Here's a demo—see 🧵 for the rest

Andy Bromberg

26,464 次观看 • 1 年前

Used Karpathy's LLM Wiki to turn Claude Code into a self-evolving system. Fed it my frontend & CRM sources, and it built a fully updating dashboard with Shadcn packages and constantly improving and linking everything for smarter outputs! Full setup video:

Used Karpathy's LLM Wiki to turn Claude Code into a self-evolving system. Fed it my frontend & CRM sources, and it built a fully updating dashboard with Shadcn packages and constantly improving and linking everything for smarter outputs! Full setup video:

WorldofAI

133,594 次观看 • 6 个月前

CLAUDE COWORK TOUR: I built my own rules and memory system (harness) so that Claude Cowork remembers everything I do. In this video I show you the specific folder structures, files, and why I did it this way. I use this same system for all 3 of my companies.

CLAUDE COWORK TOUR: I built my own rules and memory system (harness) so that Claude Cowork remembers everything I do. In this video I show you the specific folder structures, files, and why I did it this way. I use this same system for all 3 of my companies.

Justin Brooke

26,611 次观看 • 6 个月前

He is 18 built an LLM and sold it to Anthropic for $3.2M - and at Stanford shared how to repeat it from scratch: 00:30 - how to build an LLM worth $100k in one evening 21:34 - how one LLM controls 1000 processes at the same time 43:47 - from first prompt to $3.2M from Anthropic after watching I built my first LLM in 60 minutes- a week later Anthropic offered me $100k for the system. save & watch - the article below shows how to build an LLM and sell it to Anthropic for $3.2M from scratch

He is 18 built an LLM and sold it to Anthropic for $3.2M - and at Stanford shared how to repeat it from scratch: 00:30 - how to build an LLM worth $100k in one evening 21:34 - how one LLM controls 1000 processes at the same time 43:47 - from first prompt to $3.2M from Anthropic after watching I built my first LLM in 60 minutes- a week later Anthropic offered me $100k for the system. save & watch - the article below shows how to build an LLM and sell it to Anthropic for $3.2M from scratch

Noisy

46,708 次观看 • 1 个月前

The whole session ranks among my favorite David talks. Built friction out of the system not into the system and the XRP Ledger is built for business!

The whole session ranks among my favorite David talks. Built friction out of the system not into the system and the XRP Ledger is built for business!

Vet

35,717 次观看 • 1 年前

SCIENTISTS TURN LIGHT INTO A MEMORY SYSTEM Scientists built a tiny photonic chip and sent identical photons through it, expecting messy quantum behavior. Instead, the light started acting like a memory system, forming patterns that match a classic AI model used for associative memory. In simple terms, the interference between photons did the math needed to store and recall patterns. No brain, no computer, just light physics doing the computing. Source: NewsForce

SCIENTISTS TURN LIGHT INTO A MEMORY SYSTEM Scientists built a tiny photonic chip and sent identical photons through it, expecting messy quantum behavior. Instead, the light started acting like a memory system, forming patterns that match a classic AI model used for associative memory. In simple terms, the interference between photons did the math needed to store and recall patterns. No brain, no computer, just light physics doing the computing. Source: NewsForce

NewsForce

29,578 次观看 • 7 个月前

LLM + Memory + Planning + Tools = Agents 🤖 Last month, Job and I discussed how generative AI is shifting how companies offer customer support. How can we add more layers to our RAG apps to make it more agentic? LLM: Large language model alone Memory: Short-term and long-term memory Planning: Breaking down questions into sub-questions, plan the steps needed before taking action, etc. Tools: Calculator, code executor, weather API, etc. I hope you find this small clip interesting! The link to the full recording is below.

LLM + Memory + Planning + Tools = Agents 🤖 Last month, Job and I discussed how generative AI is shifting how companies offer customer support. How can we add more layers to our RAG apps to make it more agentic? LLM: Large language model alone Memory: Short-term and long-term memory Planning: Breaking down questions into sub-questions, plan the steps needed before taking action, etc. Tools: Calculator, code executor, weather API, etc. I hope you find this small clip interesting! The link to the full recording is below.

Erika Shorten

32,443 次观看 • 2 年前

Fellas, stop guessing. I built a system that turns content into income, and I’m teaching you step-by-step. This ain’t luck, it’s strategy!

Fellas, stop guessing. I built a system that turns content into income, and I’m teaching you step-by-step. This ain’t luck, it’s strategy!

Tysin Black 2.0

11,485 次观看 • 1 年前

Introducing Brain in Computer. Brain is a continuously learning memory system. Every task on Computer plugs into a context graph built by Brain. It makes Computer more stateful with every run. Available as a research preview for all Perplexity Max subscribers.

Introducing Brain in Computer. Brain is a continuously learning memory system. Every task on Computer plugs into a context graph built by Brain. It makes Computer more stateful with every run. Available as a research preview for all Perplexity Max subscribers.

Perplexity

612,499 次观看 • 4 个月前

I built the best memory for voice agents while sponsoring Y Combinator hackathon. Introducing: Voice AI + Memory! We built a deep supermemory integration for Pipecat AI, so now even your agents have great memory With user profiles, these are almost-instant latency as well :)

I built the best memory for voice agents while sponsoring Y Combinator hackathon. Introducing: Voice AI + Memory! We built a deep supermemory integration for Pipecat AI, so now even your agents have great memory With user profiles, these are almost-instant latency as well :)

Dhravya Shah

83,511 次观看 • 7 个月前

How I Built Grey Matter, A Procedural Data System in Blender

How I Built Grey Matter, A Procedural Data System in Blender

Sam Hox

84,279 次观看 • 1 年前