Harshveer Singh

.. AI Researcher ..

I truly believe that degrees of consciousness are hierarchical at all levels and originate due to superposition collapses.

Views my own.

[Harshveer Singh @stringharsh]

At any T=t, quantum states in the brain, call them "prior", correspond to an existing abstract universe somewhere...
we "think" and set a specific pattern of entangled qubits, call it "evidence", which corresponds to entangled black holes forming in that inner universe...

[空言 @__soragoto__]

1990s：我们用上了 B+ 树和倒排索引，世界上已经没有什么几秒钟查不出来的东西了。

2010s：我们推广了分布式全文搜索引擎，几十亿条日志毫秒级命中。

2025：我们把所有文档切成稀碎的 chunk，烧掉了不知道多少算力把它们向量化后扔进向量数据库，最后把搜出来的垃圾拼成几万字 prompt 塞进大模型，然后开始祈祷它别出现幻觉。

2026：大模型告诉我们，这个 find，还有这个 grep 工具真好用。

[Harshveer Singh @stringharsh]

Jev + TimesFM-3 could ressurect Jim Simmons

[Brexy @brexy_ai]

Security matters more when AI becomes part of financial work.

In our latest blog, we share what this milestone means for Brexy and our commitment to building secure, trustworthy AI for financial services.

Read the blog → https://t.co/ofhKyo7SOv

#Brexy #SOC2 #AISecurity

[Physics In History @PhysInHistory]

Grigori Perelman ✍️

[Pata van Goon @basedalexandoor]

OpenAI: We're gonna destroy your career

Tristan: "I'm an academic"

[Brexy @brexy_ai]

What if financial research took minutes, not hours?

Brexy helps finance teams go from information → insight → decision faster.

Analyze documents. Find insights. Cut repetitive research.

Research faster. Analyze deeper. Decide better.

https://t.co/cD20hYpJsw

#AI #Finance

[Harshveer Singh @stringharsh]

@OpenAI You didnt solve it come on, you stole it off of your date from that mathematician, you just out-conputed him. Have some shame posting this on X

[Tim Rocktäschel @_rockt]

We are stuck in a massive local optimum.

[How To Prompt @HowToPrompt__]

Plato was actually right..

Researchers proved every LLM on earth is converging on the exact same "universal geometry" of meaning.

They built a method that can translate between ANY model's embeddings without ever seeing the original text or using paired data.

different architectures, different training sets, different parameter counts.. it doesn't matter.

Until now, every AI model has lived in its own isolated mathematical universe.

An embedding vector from Claude meant nothing to GPT, and a vector from Llama meant nothing to Gemini. They spoke entirely different geometric languages.

To bridge them, you always needed paired datasets, complex encoders, or heavy fine-tuning.

Then researchers dropped a bombshell paper.

They built a system that can translate between any model's embeddings without ever seeing the original text, without encoders, and without a single pair of matching data.

How?

Because the geometry is already there.

Different models, built by different companies, with totally different architectures, parameter counts, and training data, are all naturally drifting toward the exact same underlying latent structure of human meaning.

The Platonic Representation Hypothesis isn't just a theory anymore. It’s a mathematical reality.

They built an unsupervised method that maps an unknown embedding from one model straight into a universal representation space, matching text vectors across different models with shockingly high precision.

But here is the dark side nobody is talking about.

If meaning has a universal geometry, and vectors can be freely translated across models without the original text or encoders...

Vector databases are wide open.

An adversary with access only to a company's stored embedding vectors can translate them, invert them, and extract sensitive internal documents, personal data, and proprietary codebases without ever hacking the model itself.

[Harshveer Singh @stringharsh]

@GonnabeNikhil Mid, no doubt why they just piggybacked on deepseek's work...Referring to copying as research is not new for Indian companies.

[Harshveer Singh @stringharsh]

@Dank_jetha 😭😭😭🤣🤣🤣🤣🤣🤣🤣🤣🤣

[Harshveer Singh @stringharsh]

@tiannah_tee Bruh, look whos talking about infidelity, changin bros like clothes

[Harshveer Singh @stringharsh]

@0xEronn its Geoffrey not Jeff

[Brexy @brexy_ai]

Your office may be empty at 2:13 AM. Your institution isn't.

Brexy helps capital markets teams automate workflows, monitor information and keep institutional intelligence moving around the clock.

AI Operating System for Capital Markets.

Book a demo → https://t.co/cD20hYpJsw

[Brexy @brexy_ai]

Earnings analysis shouldn’t mean hours of digging.
AI can help asset managers turn earnings calls, filings, and market data into actionable insights, faster.

Here’s how AI is reshaping earnings analysis ↓

https://t.co/uRot24QIzq

#AI #AssetManagement #FinTech

[Brexy @brexy_ai]

Every deal team wants more hours in the day.

The better solution? Eliminate the work that shouldn't take hours.

Here are 9 ways Brexy gives deal teams their time back.

Less searching. More thinking.

Book a demo → https://t.co/I8QiNDnE1J

#AI #CapitalMarkets

[Brexy @brexy_ai]

Every deal team knows the pain: the data exists, but it’s buried everywhere.

Brexy helps turn scattered information into actionable intelligence so teams spend less time searching and more time closing.

🎥 See it in action: https://t.co/I8QiNDnE1J

#AI #CapitalMarkets #Brexy

[Harshveer Singh @stringharsh]

@sama Is this supposed to be tuff in 2026?

[Brexy @brexy_ai]

How many hours does your team spend searching documents before making a decision?

What if the answer was minutes instead of hours?

Watch how Brexy helps capital markets teams research faster, analyze smarter, and execute deals with AI.

Book a demo: https://t.co/Cq22vx6QiH

#AI

[Harshveer Singh @stringharsh]

@alokbishoyi97 So its all agentic post training and inference optimization work?