Introducing ScaleDown
15x cheaper. 63x faster. 5.1% more accurate than GPT-5.4 Mini.
Task-specific SLMs for the 70-80% of AI workloads that don't need a frontier model.
From NeurIPS '25 to https://t.co/imZLOYCcTA
Three routing innovations this week.
@FactoryAI routes the model: frontier quality, 25% lower cost.
@Harvey + @FireworksAI_HQ route the advisor: open model primary; frontier model selectively invoked.
Rekursor routes inspectable, customer-owned skills selected at runtime, with a new primitive that shows 100% top-pick correctness across an 18 → 504 skill library while RAG collapses past ~200 candidates.
Different layers. Same wave. Routing is becoming the architecture.
https://t.co/n9piULQ2XT
“The biggest reason this is so effective is that no one has done this in SF…. In order to go viral with paid advertising, you have to try something new. Take risk. Be creative. Otherwise you're paying a premium just to be like everyone else with a billboard on a highway.”
Levels of AI app layer companies:
1. Use frontier models to get to PMF
2. Start mixing open source models to get cost under control
3. Fine tune oss model and use it with frontier advisor
4. Train your own foundational model.
Cursor is the only company I know attempting #4 only bc they got capital and expertise from xAI.
Frontier lab IPO rush is making a lot more sense seeing something like this. The greatest competition was always with open weights! Multi agent systems using a smarter “advisor” model to drive workers is something we’ve had success with @finchlegal already. Very bullish on the future of “build / host your own workflow specific agents”
This quote from Steve Jobs has always resonated with me. An important reminder as we write the chapters of our life's book.
"Your time is limited, so don't waste it living someone else's life. Don't be trapped by dogma, which is living with the results of other people's thinking. Don't let the noise of others' opinions drown out your own inner voice. And most important, have the courage to follow your heart and intuition. They somehow already know what you truly want to become. Everything else is secondary."
I think the biggest thing YC changed for me was reducing noise.
Before YC, startup advice on the internet felt very random.
Everyone sounded smart.
Everyone had opinions.
A lot of it was bluff.
YC made startups feel much more grounded:
ARR, retention, distribution, market size, growth.
There’s a much more standardized way to evaluate whether a business is actually working.
At the end of the day, the best founders I’ve met are usually just trying to build real businesses with real demand.
Not optimize for appearances.
There's a new playbook being written right now for running an AI-native service business, and @charliewarren is at the frontier. Great advice here for how to do it right.
This is the new Intel.
Focused. Executing. Building across the full AI stack.
With new products, deeper partnerships, and purpose-built solutions for the AI era.
https://t.co/px7Ho5MTgf
Tokens/inferencing/intelligence becomes just another resource to manage just like talent, money or raw materials.
We’ll need new tools ti mange it, measure it, procure it, attribute and monetize
@patrick_oshag this interview and Stanford course by @apoorv03 https://t.co/CRr6nyBtSL have been the best resources on economics resources on economics of AI
@patrick_oshag this interview and Stanford course by @apoorv03 https://t.co/CRr6nyBtSL have been the best resources on economics resources on economics of AI
@arram we need techno optimist manifesto for AI where we argue that every company in the future will be an AI company, and benefits will be accrued not just two 2 models companies.
this is the only other viable competing narrative to communist-UBI vision Dario painted in his essays.