Sophy Yan

All things AI | VC & Entrepreneur | Partner@NexParadigm Capital *Opinions here are my own

[Sophy Yan @sophyyyyan]

Regarding the Turing Test, I believe it is designed to measure intelligence rather than sentience. If a machine can convincingly imitate a human, it has effectively outsmarted you, assuming you are intelligent(?!🤪)
credit:#midjourney #GenerativeAI #AIArtworks

[Sophy Yan @sophyyyyan]

@TrustSpooky @ryanels @TrustGraphAI

[Sophy Yan @sophyyyyan]

Can machines think? First prove that humans can.

#AIandThought #LLMs#HumanInTheLoop 🤖🧠

[Sophy Yan @sophyyyyan]

@TrustGraphAI @TrustSpooky @cybermaggedon 🎉 1,000 stars! Congrats to the TrustGraph team and this amazing community—turning messy enterprise context into open, usable, production-ready graphs. 🚀

#OpenSource #ContextGraphs #AI

[Sophy Yan @sophyyyyan]

@the_judge1111 Writes policies for every self-driving car $1B ARR instant!

[Sophy Yan @sophyyyyan]

@nikhilvnamburi Insures every home in flood zone in Florida $2B ARR

[Sophy Yan @sophyyyyan]

@TrustSpooky @TrustGraphAI Congrats 🚀 Context graphs are clearly striking a nerve—this feels like a real inflection point for AI moving from clever demos to grounded, decision-aware systems. Seeing TrustGraph trending on GitHub is well deserved. Excited to see what builders unlock with this foundation.

[Sophy Yan @sophyyyyan]

@bbldrizzedup @TrustSpooky Generalizing is hard, but useful in domain-specific languages where things are well-defined, such as legal, finance, medical etc

[Sophy Yan @sophyyyyan]

@cummutacation @TrustSpooky Bigger buckets don’t fix messy pipelines

[Sophy Yan @sophyyyyan]

@TrustGraphAI On-prem SLMs with context graphs: fewer tokens, fewer leaks, fewer invoices 💸💸💸
#AI #LLMs #ContextGraph #KnowledgeGraphs #Graphrag #GenerativeAI

[Daniel Davis @TrustSpooky]

https://t.co/qmEcAa9amh

[Sophy Yan @sophyyyyan]

@DisruptConcept LLMs don’t suffer from lack of intelligence — they suffer from lack of context. Graphs feel like the missing connective tissue #AI #LLMs #ContextGraph #KnowledgeGraphs #GraphRag #GenerativeAI #TrustGraph

[Sophy Yan @sophyyyyan]

@TrustGraphAI @huntbase_io Love seeing events like this at SF@Techweek_ Agentic AI is the next frontier in SecOps🔥

[Sophy Yan @sophyyyyan]

@TrustGraphAI AI can’t run on maybe.
It runs on trusted, real-time data.
TrustGraph is making that real at #DataStreamingSummit2025.
📍 San Francisco

[Sophy Yan @sophyyyyan]

@godofprompt @OpenAI @Oracle @nvidia The interplay of subsidized GPUs, cloud credits, and artificially recycled demand within a closed looped ecosystem creates a facade of exponential growth, while the bulk of the money generated currently in AI are just from financial engineering in the public markets

[Sophy Yan @sophyyyyan]

@B_AI_S Addressing determinism is meaningful only if it’s enforced across the entire tech stack; the paper from Thinking Machines constrains variability in very narrow contexts —batching. See this article for more insight: https://t.co/61SzaW5q44

[Daniel Davis @TrustSpooky]

Thinking Machines, one of the OpenAI spinoff labs just released a paper on LLM determinism.

And I'm a bit perplexed at the retconning of the term "determinism"...

In another entry in, “I read an AI paper so you don’t have to”, Thinking Machines just released, “Defeating Nondeterminism in LLM Inference”. Hmm. Are LLMs still using neural networks? Check. And they’re arguing it can be made deterministic? Insert your favorite speechless face meme here.

[Sophy Yan @sophyyyyan]

@gm8xx8 Addressing determinism is meaningful only if it’s enforced across the entire tech stack; the paper from Thinking Machines constrains variability in very narrow contexts —batching. See this article for more insight: https://t.co/61SzaW5q44

[Daniel Davis @TrustSpooky]

Thinking Machines, one of the OpenAI spinoff labs just released a paper on LLM determinism.

And I'm a bit perplexed at the retconning of the term "determinism"...

In another entry in, “I read an AI paper so you don’t have to”, Thinking Machines just released, “Defeating Nondeterminism in LLM Inference”. Hmm. Are LLMs still using neural networks? Check. And they’re arguing it can be made deterministic? Insert your favorite speechless face meme here.

[Sophy Yan @sophyyyyan]

@wesroth Addressing determinism is meaningful only if it’s enforced across the entire tech stack; the paper from Thinking Machines constrains variability in very narrow contexts —batching. See this article for more insight: https://t.co/61SzaW5q44

[Daniel Davis @TrustSpooky]

Thinking Machines, one of the OpenAI spinoff labs just released a paper on LLM determinism.

And I'm a bit perplexed at the retconning of the term "determinism"...

In another entry in, “I read an AI paper so you don’t have to”, Thinking Machines just released, “Defeating Nondeterminism in LLM Inference”. Hmm. Are LLMs still using neural networks? Check. And they’re arguing it can be made deterministic? Insert your favorite speechless face meme here.

[Sophy Yan @sophyyyyan]

@HeMuyu0327 @thinkymachines Addressing determinism is meaningful only if it’s enforced across the entire tech stack; the paper from Thinking Machines constrains variability in very narrow contexts —batching. See this article for more insight: https://t.co/61SzaW5q44

[Daniel Davis @TrustSpooky]

Thinking Machines, one of the OpenAI spinoff labs just released a paper on LLM determinism.

And I'm a bit perplexed at the retconning of the term "determinism"...

In another entry in, “I read an AI paper so you don’t have to”, Thinking Machines just released, “Defeating Nondeterminism in LLM Inference”. Hmm. Are LLMs still using neural networks? Check. And they’re arguing it can be made deterministic? Insert your favorite speechless face meme here.

[Sophy Yan @sophyyyyan]

@arjunkocher @cHHillee Addressing determinism is meaningful only if it’s enforced across the entire tech stack; the paper from Thinking Machines constrains variability in very narrow contexts —batching. See this article for more insight: https://t.co/61SzaW5q44

[Daniel Davis @TrustSpooky]

Thinking Machines, one of the OpenAI spinoff labs just released a paper on LLM determinism.

And I'm a bit perplexed at the retconning of the term "determinism"...

In another entry in, “I read an AI paper so you don’t have to”, Thinking Machines just released, “Defeating Nondeterminism in LLM Inference”. Hmm. Are LLMs still using neural networks? Check. And they’re arguing it can be made deterministic? Insert your favorite speechless face meme here.

[Sophy Yan @sophyyyyan]

@mediocrelychee Addressing determinism is meaningful only if it’s enforced across the entire tech stack; the paper from Thinking Machines constrains variability in very narrow contexts —batching. See this article for more insight: https://t.co/61SzaW5q44

[Daniel Davis @TrustSpooky]

Thinking Machines, one of the OpenAI spinoff labs just released a paper on LLM determinism.

And I'm a bit perplexed at the retconning of the term "determinism"...

In another entry in, “I read an AI paper so you don’t have to”, Thinking Machines just released, “Defeating Nondeterminism in LLM Inference”. Hmm. Are LLMs still using neural networks? Check. And they’re arguing it can be made deterministic? Insert your favorite speechless face meme here.