Good1

Angel Investor | 2x start up founder | Silicon Valley | Drinking the coolaid!

[Good1 @MLMusings]

@AnthropicAI Is this the best thing that could have happened to Anthropic pre IPO, demand for Fable 5 will go through the roof.

[Good1 @MLMusings]

@deedydas Strong performance on Finance data

[Good1 @MLMusings]

@yashvarpatel @deedydas That’s great!

[Good1 @MLMusings]

🧵4/n VC, PE and Public equity fund the growth of American businesses, leading to heavy dilution of the founder's stake by the time a company reaches Fortune 500

Conversely, Indian enterprises historically scaled using bank debt, retained earnings, or pooled family wealth, leading to founding families retaining majority ownership and operational control

[Good1 @MLMusings]

🧵1/n 51% of Indian business are family owned, compared with 6% in the US
Source: The Economist

[Good1 @MLMusings]

🧵3/n - Tarun Khanna's Institutional Voids - Developed markets like the US have mature specialized intermediaries like
- VC/PE
- Credit rating agencies
- Strong judicial systems

In emerging market the ease of doing business is low cause theses intermediaries aren't fully mature yet

[Good1 @MLMusings]

🧵2/n Key reasons behind this -

• Inheritance or Estate Tax - 40% in US vs 0% in India

This 40% US estate tax bill has to be paid within 9 months of the owner's death, leaving the heirs with a massive liquidity crisis

India's 85% estate duty was abolished in 1985, and multi billion dollar conglomerates can be passed down tax free

[Deedy @deedydas]

Claude Mythos just obliterated every single benchmark in AI. I can't believe what I'm reading.

[Good1 @MLMusings]

@deedydas @Scobleizer Feels like an O1 moment

[Good1 @MLMusings]

@egradman @monologue @danshipper @bran_don_gell @egradman what cool workflow do you have hooked up to your monologue pedal?

[Good1 @MLMusings]

@BlackHC 💯

Unless it’s a choice between a top university and a once in a lifetime company

[Good1 @MLMusings]

https://t.co/PbFH2G1RX9

[Good1 @MLMusings]

"Meta freezes hiring for its Super Intelligence lab"

Some think its cause of AI bubble fears,

however it could also be a pause to structure the investment Meta has already made. They hired and grew the lab at an unprecedented rate, along with managing a Scale AI acquisition.

Situations like this demand extensive housekeeping

[Denny Zhou @denny_zhou]

Slides for my lecture “LLM Reasoning” at Stanford CS 25: https://t.co/eApGUHyIDo

Key points:
1. Reasoning in LLMs simply means generating a sequence of intermediate tokens before producing the final answer. Whether this resembles human reasoning is irrelevant. The crucial insight is that transformer models can become nearly arbitrarily powerful by generating many intermediate tokens, without the need of scaling the model size (https://t.co/HO2seV6vVl).

2. Pretrained models, even without any fine-tuning, are capable of reasoning. The challenge is that reasoning-based outputs often don’t appear at the top of the output distribution, so standard greedy decoding fails to surface them (https://t.co/75h2QQzT9M)

3. Prompting techniques (e.g., chain-of-thought prompting or "let’s think step by step") and supervised finetuning were commonly used to elicit reasoning. Now, RL finetuning has emerged as the most powerful method. This trick was independently discovered by several labs. At Google, credit goes to Jonathan Lai on my team. Based on our theory ( see point 1), scaling RL should focus on generating long responses rather than something else.

4. LLM reasoning can be hugely improved by generating multiple responses and then aggregating them, rather than relying on a single response (https://t.co/BA5MUzg3PR).

[elvis @omarsar0]

Comprehensive Taxonomy of LLM Hallucinations

Nice report covering common hallucinations, root causes, and mitigation strategies.

[Good1 @MLMusings]

@rasbt @natolambert Me too

[Good1 @MLMusings]

@jasonlk Moats have been breached; tech is in a tough spot now.

It won’t be long before Anthropic, Google, XAI, and Meta will have models performing at or better levels than GPT-5.

Unfortunately, everyone has the same idea, and gone are the days when there was only one Stripe or Airbnb.

[Good1 @MLMusings]

Leave of absence is a much better, low risk exploration option. If things don’t workout, go back and complete the degree.

[Paul Graham @paulg]

Don't drop out of college to start or work for a startup. There will be other (and probably better) startup opportunities, but you can't get your college years back.

[Good1 @MLMusings]

Table stakes imho

[🚨 AI News | TestingCatalog @testingcatalog]

Anthropic is testing Ghost Chats on Claude for iOS (a button on the top right).

Incognito 👻