Jay Lye
@jnlsk.bsky.social
Fullstack AI Engineer @ Stealth, building agentic systems, crafting AI experiences
Google Startup Cloud Program participant. Top global rankings in Nvidia's GenAI contests: LangChain (Jun '24, 14th-113th) & LlamaIndex (Nov '24, 17th-126th).
Google Startup Cloud Program participant. Top global rankings in Nvidia's GenAI contests: LangChain (Jun '24, 14th-113th) & LlamaIndex (Nov '24, 17th-126th).
Pinned
Revealing the Hidden Economics of Open Models in the AI Era
Frank Nagle, research scientist at MIT and chief economist of the Linux Foundation, discusses his recent paper on the economic role of open models.
www.linuxfoundation.org
Opinion: every closed-model provider is feeling the costs/performance pressure from open models
#linux #opensource #language-models #llm #ai #agentic-system
#linux #opensource #language-models #llm #ai #agentic-system
Reposted by Jay Lye
We present Olmo 3, our next family of fully open, leading language models.
This family of 7B and 32B models represents:
1. The best 32B base model.
2. The best 7B Western thinking & instruct models.
3. The first 32B (or larger) fully open reasoning model.
This family of 7B and 32B models represents:
1. The best 32B base model.
2. The best 7B Western thinking & instruct models.
3. The first 32B (or larger) fully open reasoning model.
November 20, 2025 at 2:32 PM
We present Olmo 3, our next family of fully open, leading language models.
This family of 7B and 32B models represents:
1. The best 32B base model.
2. The best 7B Western thinking & instruct models.
3. The first 32B (or larger) fully open reasoning model.
This family of 7B and 32B models represents:
1. The best 32B base model.
2. The best 7B Western thinking & instruct models.
3. The first 32B (or larger) fully open reasoning model.
Opinion: every closed-model provider is feeling the costs/performance pressure from open models
#linux #opensource #language-models #llm #ai #agentic-system
#linux #opensource #language-models #llm #ai #agentic-system
Revealing the Hidden Economics of Open Models in the AI Era
Frank Nagle, research scientist at MIT and chief economist of the Linux Foundation, discusses his recent paper on the economic role of open models.
www.linuxfoundation.org
November 20, 2025 at 6:11 AM
Opinion: every closed-model provider is feeling the costs/performance pressure from open models
#linux #opensource #language-models #llm #ai #agentic-system
#linux #opensource #language-models #llm #ai #agentic-system
Reposted by Jay Lye
Extremely upset that a throw-away XKCD joke somehow became the organizing principle for the Internet.
November 20, 2025 at 2:31 AM
Extremely upset that a throw-away XKCD joke somehow became the organizing principle for the Internet.
Opinion: once again Cloudflare delivering a masterclass in corporate communication
Misconfigurations are hard to avoid, a cascading failure ensures when it does happen.
#cloudflare #cybersecurity #internet
Misconfigurations are hard to avoid, a cascading failure ensures when it does happen.
#cloudflare #cybersecurity #internet
Cloudflare outage on November 18, 2025
Cloudflare suffered a service outage on November 18, 2025. The outage was triggered by a bug in generation logic for a Bot Management feature file causing many Cloudflare services to be affected.
blog.cloudflare.com
November 19, 2025 at 2:40 AM
Opinion: once again Cloudflare delivering a masterclass in corporate communication
Misconfigurations are hard to avoid, a cascading failure ensures when it does happen.
#cloudflare #cybersecurity #internet
Misconfigurations are hard to avoid, a cascading failure ensures when it does happen.
#cloudflare #cybersecurity #internet
Opinion: there's an ongoing ancient battle between human rationality vs emotion, my bet's on rationality in the long term.
November 10, 2025 at 8:52 AM
Opinion: there's an ongoing ancient battle between human rationality vs emotion, my bet's on rationality in the long term.
Opinion: language models are rapidly depreciating assets
#mistralai #llm #artificialintelligence #agentic-system #ai
#mistralai #llm #artificialintelligence #agentic-system #ai
Magistral | Mistral AI
Stands to reason.
mistral.ai
September 18, 2025 at 6:12 AM
Opinion: language models are rapidly depreciating assets
#mistralai #llm #artificialintelligence #agentic-system #ai
#mistralai #llm #artificialintelligence #agentic-system #ai
Opinion: even multi-billion companies can't ignore the prospects of gaining market share with free products and selling you the ecosystem
#openai #llm #agentic-system #artificial-intelligence #ai
#openai #llm #agentic-system #artificial-intelligence #ai
Introducing gpt-oss
gpt-oss-120b and gpt-oss-20b push the frontier of open-weight reasoning models
openai.com
August 6, 2025 at 6:45 AM
Opinion: even multi-billion companies can't ignore the prospects of gaining market share with free products and selling you the ecosystem
#openai #llm #agentic-system #artificial-intelligence #ai
#openai #llm #agentic-system #artificial-intelligence #ai
Opinion: MMR reranking is great for returning diverse results especially in datasets with overlapping entries
#qdrant #rag #retrieval-augmented-generation #ai #artificialintelligence #llm
#qdrant #rag #retrieval-augmented-generation #ai #artificialintelligence #llm
Qdrant 1.15 - Smarter Quantization & better Text Filtering - Qdrant
Qdrant v1.15 release presents new Quantization Features, advanced Full-Text filtering and a bunch of performance optimizations
qdrant.tech
July 19, 2025 at 1:57 AM
Opinion: MMR reranking is great for returning diverse results especially in datasets with overlapping entries
#qdrant #rag #retrieval-augmented-generation #ai #artificialintelligence #llm
#qdrant #rag #retrieval-augmented-generation #ai #artificialintelligence #llm
Opinion: every public library/institution should be running their own MCP servers for kids, students, professionals and retirees in the 4th Industrial Revolution
#mcp #modelcontextprotocol #llm #agentic-system #language-model #ai #artitificial-intelligence
#mcp #modelcontextprotocol #llm #agentic-system #language-model #ai #artitificial-intelligence
July 7, 2025 at 10:28 AM
Opinion: every public library/institution should be running their own MCP servers for kids, students, professionals and retirees in the 4th Industrial Revolution
#mcp #modelcontextprotocol #llm #agentic-system #language-model #ai #artitificial-intelligence
#mcp #modelcontextprotocol #llm #agentic-system #language-model #ai #artitificial-intelligence
Opinion: cost of hosting DeepSeek in 2025 on 96 H100 ranges from USD$192/hr to USD$1181/hr(excl. taxes) depending on configuration and location
#deepseek #sglang #agentic-system #llm #artificial-intelligence
#deepseek #sglang #agentic-system #llm #artificial-intelligence
Deploying DeepSeek with PD Disaggregation and Large-Scale Expert Parallelism on 96 H100 GPUs | LMSYS Org
<p>DeepSeek is a popular open-source large language model (LLM) praised for its strong performance. However, its large size and unique architecture, which us...
lmsys.org
June 24, 2025 at 10:56 AM
Opinion: cost of hosting DeepSeek in 2025 on 96 H100 ranges from USD$192/hr to USD$1181/hr(excl. taxes) depending on configuration and location
#deepseek #sglang #agentic-system #llm #artificial-intelligence
#deepseek #sglang #agentic-system #llm #artificial-intelligence
Opinion: future AI systems will likely be multi-agents as they tackle increasingly complex tasks, translating context and instructions into economic value
#anthropic #agentic-system #llm #artificial-intelligence #ai
#anthropic #agentic-system #llm #artificial-intelligence #ai
How we built our multi-agent research system
On the the engineering challenges and lessons learned from building Claude's Research system
www.anthropic.com
June 18, 2025 at 1:59 AM
Opinion: future AI systems will likely be multi-agents as they tackle increasingly complex tasks, translating context and instructions into economic value
#anthropic #agentic-system #llm #artificial-intelligence #ai
#anthropic #agentic-system #llm #artificial-intelligence #ai
Reposted by Jay Lye
If you use "AI agents" (LLMs calling tools in a loop) you need to be aware of the Lethal Trifecta
Combine access to private data, exposure to untrusted content and the ability to externally communicate and an attacker can trick the system into stealing your data simonwillison.net/2025/Jun/16/...
Combine access to private data, exposure to untrusted content and the ability to externally communicate and an attacker can trick the system into stealing your data simonwillison.net/2025/Jun/16/...
The lethal trifecta for AI agents: private data, untrusted content, and external communication
If you are a user of LLM systems that use tools (you can call them “AI agents” if you like) it is critically important that you understand the risk of …
simonwillison.net
June 16, 2025 at 1:21 PM
If you use "AI agents" (LLMs calling tools in a loop) you need to be aware of the Lethal Trifecta
Combine access to private data, exposure to untrusted content and the ability to externally communicate and an attacker can trick the system into stealing your data simonwillison.net/2025/Jun/16/...
Combine access to private data, exposure to untrusted content and the ability to externally communicate and an attacker can trick the system into stealing your data simonwillison.net/2025/Jun/16/...
Reposted by Jay Lye
Japan-UK, too. #FusionEnergy asia.nikkei.com/Business/Tec...
Japan and UK to sign memorandum for nuclear fusion collaboration
Tokyo seeks pioneering role in new power generation by the 2030s
asia.nikkei.com
June 16, 2025 at 2:23 PM
Japan-UK, too. #FusionEnergy asia.nikkei.com/Business/Tec...
Opinion: Cloudflare should be giving a masterclass in post-mortem reports. The tone, clarity, timeline and remediation included should be the minimal standard in 21st century.
#cloudflare #data-center #cloud-computing #web #saas #internet
#cloudflare #data-center #cloud-computing #web #saas #internet
Cloudflare service outage June 12, 2025
Today, June 12, 2025, Cloudflare suffered a significant service outage that affected a large set of our critical services, including Workers KV, WARP, Access, Gateway, Images, Stream, Workers AI, Turn...
blog.cloudflare.com
June 14, 2025 at 12:28 AM
Opinion: Cloudflare should be giving a masterclass in post-mortem reports. The tone, clarity, timeline and remediation included should be the minimal standard in 21st century.
#cloudflare #data-center #cloud-computing #web #saas #internet
#cloudflare #data-center #cloud-computing #web #saas #internet
Opinion: SOTA VLMs are liars, and they lie confidently by regurgitating memorized training data
#vlm #vision-language-model #ai #artificial-intelligence #agentic-system #llm #language-model #visual-analysis #visual-reasoning #cognitive-science #neuroscience
#vlm #vision-language-model #ai #artificial-intelligence #agentic-system #llm #language-model #visual-analysis #visual-reasoning #cognitive-science #neuroscience
Vision Language Models are Biased
Vision Language Models are Biased: VLMs fail on simple counting tasks when familiar objects are subtly modified
vlmsarebiased.github.io
June 4, 2025 at 5:36 AM
Opinion: SOTA VLMs are liars, and they lie confidently by regurgitating memorized training data
#vlm #vision-language-model #ai #artificial-intelligence #agentic-system #llm #language-model #visual-analysis #visual-reasoning #cognitive-science #neuroscience
#vlm #vision-language-model #ai #artificial-intelligence #agentic-system #llm #language-model #visual-analysis #visual-reasoning #cognitive-science #neuroscience
Opinion: one step closer to self-improving AI
#sakanaai #agentic-system #llm #evolution #biology #swe-bench
#software-engineering #computer-science
#sakanaai #agentic-system #llm #evolution #biology #swe-bench
#software-engineering #computer-science
Sakana AI
The Darwin Gödel Machine: AI that improves itself by rewriting its own code
sakana.ai
June 4, 2025 at 5:27 AM
Opinion: one step closer to self-improving AI
#sakanaai #agentic-system #llm #evolution #biology #swe-bench
#software-engineering #computer-science
#sakanaai #agentic-system #llm #evolution #biology #swe-bench
#software-engineering #computer-science
Opinion: closed source models feeling the heat
#qwen #alibaba #vllm #ai #artificial-intelligence #agentic-ai
#qwen #alibaba #vllm #ai #artificial-intelligence #agentic-ai
Qwen3: Think Deeper, Act Faster
QWEN CHAT GitHub Hugging Face ModelScope Kaggle DEMO DISCORD
Introduction Today, we are excited to announce the release of Qwen3, the latest addition to the Qwen family of large language models. Our f...
qwenlm.github.io
April 29, 2025 at 4:56 AM
Opinion: closed source models feeling the heat
#qwen #alibaba #vllm #ai #artificial-intelligence #agentic-ai
#qwen #alibaba #vllm #ai #artificial-intelligence #agentic-ai
Opinion: even those with a high degree of patience will eventually run out of tolerance when web scraping becomes unmanageable
#wikipedia #ai #artificial-intelligence #agentic-system #kaggle #google
#wikipedia #ai #artificial-intelligence #agentic-system #kaggle #google
Wikipedia Kaggle Dataset using Structured Contents Snapshot
A structured Wikipedia dataset is now on Kaggle from Wikimedia Enterprise, built for ML workflows using the Snapshot API’s Structured Contents beta.
enterprise.wikimedia.com
April 17, 2025 at 11:46 AM
Opinion: even those with a high degree of patience will eventually run out of tolerance when web scraping becomes unmanageable
#wikipedia #ai #artificial-intelligence #agentic-system #kaggle #google
#wikipedia #ai #artificial-intelligence #agentic-system #kaggle #google
Opinion: a reasonable balance between backing new models, ease of implementation, performance, and swift experimentation, among other inference options
Satisfactory is better than perfection when delivering SOTA models to end-users.
#vllm #transformers #huggingface #ai #artificial-intelligence
Satisfactory is better than perfection when delivering SOTA models to end-users.
#vllm #transformers #huggingface #ai #artificial-intelligence
Transformers backend integration in vLLM
The Hugging Face Transformers library offers a flexible, unified interface to a vast ecosystem of model architectures. From research to fine-tuning on custom dataset, transformers is the go-to toolkit...
blog.vllm.ai
April 17, 2025 at 5:05 AM
Opinion: a reasonable balance between backing new models, ease of implementation, performance, and swift experimentation, among other inference options
Satisfactory is better than perfection when delivering SOTA models to end-users.
#vllm #transformers #huggingface #ai #artificial-intelligence
Satisfactory is better than perfection when delivering SOTA models to end-users.
#vllm #transformers #huggingface #ai #artificial-intelligence
Reposted by Jay Lye
We hope Android users have been enjoying our latest app update, which aims to provide a frictionless and smoother overall experience.
Here's how 👇
1/5
Here's how 👇
1/5
April 15, 2025 at 3:34 PM
We hope Android users have been enjoying our latest app update, which aims to provide a frictionless and smoother overall experience.
Here's how 👇
1/5
Here's how 👇
1/5
Opinion: creatives have to update their skill set and toolbox to keep up with the exponential advancements
#text-to-video #kling #ai #artificial-intelligence
#text-to-video #kling #ai #artificial-intelligence
From the ChatGPT community on Reddit: Kling's newest AI video model make it hard to notice if video is AI or not!
Explore this post and more from the ChatGPT community
www.reddit.com
April 15, 2025 at 1:32 PM
Opinion: creatives have to update their skill set and toolbox to keep up with the exponential advancements
#text-to-video #kling #ai #artificial-intelligence
#text-to-video #kling #ai #artificial-intelligence
Opinion: this is giving me "The Expanse" vibes, I welcome the Creative Age when every viewer has the opportunity to dictate the characters/arcs/endings with their favorite content
#ai #artificial-inteligence #agentic-system #genai
#ai #artificial-inteligence #agentic-system #genai
From the singularity community on Reddit: Age of Beyond - An AI Assisted short I made in 2 and a half months.
Explore this post and more from the singularity community
www.reddit.com
April 6, 2025 at 7:44 AM
Opinion: this is giving me "The Expanse" vibes, I welcome the Creative Age when every viewer has the opportunity to dictate the characters/arcs/endings with their favorite content
#ai #artificial-inteligence #agentic-system #genai
#ai #artificial-inteligence #agentic-system #genai
Opinion: interesting take on AI's trajectory in the near term
#ai #artificial-intelligence #agentic-system #llm
#ai #artificial-intelligence #agentic-system #llm
AI 2027
A research-backed AI scenario forecast.
ai-2027.com
April 4, 2025 at 5:28 AM
Opinion: interesting take on AI's trajectory in the near term
#ai #artificial-intelligence #agentic-system #llm
#ai #artificial-intelligence #agentic-system #llm
Opinion: the world is not ready for an explosion of AI-led scientific discovery by both non-experts and experts
#ai #artificial-intelligence #agentic-system #llm #digital-forensics
#ai #artificial-intelligence #agentic-system #llm #digital-forensics
AI proves that human fingerprints are not unique, upending 100 years of law enforcement
Fingerprints have long been the cornerstone of law enforcement, linking suspects to crimes. Now, AI has disrupted this 100-year-old system.
www.earth.com
April 4, 2025 at 5:12 AM
Opinion: the world is not ready for an explosion of AI-led scientific discovery by both non-experts and experts
#ai #artificial-intelligence #agentic-system #llm #digital-forensics
#ai #artificial-intelligence #agentic-system #llm #digital-forensics