This is what we learned when we built @fiddler_ai Centor models. For always on agentic monitoring, frontier models kill performance and budgets - task specific models are the only way.
โโฆ a tuned open model can match frontier quality at a fraction of the cost.โ
AI has finally breached national security intelligence thresholds - everything will be unprecedented from now on as new safety processes get built on the fly.
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: https://t.co/bwn0sximKZ
Hats off to Anthropic. If you havenโt figured it out already - intelligence is accelerating at scale and will permeate to all software.
Itโs time to build!
Fable 5 is state-of-the-art on nearly all tested benchmarks, with exceptional performance in software engineering, knowledge work, scientific research, and vision.
The longer and more complex the task, the larger Fable 5โs lead over our other models.
This is already happening - for agent observability, we saw enterprises rebel against the sticker shock of evaluating behavior with another frontier LLM. We had to build 100x cheaper task specific evaluator models to solve that.
Good take
My guess is
- demand for intelligence is near infinite
- but 80% of workloads will be running on 99% cheaper models within 12-18 months
- 20% of workloads will still run on latest gen models where IQ maxing is important (scientific breakthroughs, higher level ochestrator agents?)
- rough analogy might be what % of macbooks or gaming PCs sold have the maxed out specs for CPU/GPU, prices are falling much faster than Moore's law here though
- this leads me to think the limiting factor will be energy and compute, not better models
At Coinbase we're working hard on routing prompts to cheaper models where appropriate, and in some cases have been able to keep costs roughly flat, while token usage continues to grow exponentially.
This is the tough lesson that a lot of people are learning the hard way
AI might have made building apps a lot easier, but it also set the barrier to entry at zero
Because anyone can do it, there is no moat left
The only edge left in the future will be sales and marketing
The Monaco Plane: Economics and Impact.
If you've been in SF over the last 10 days, you've likely seen the Monaco Plane flying around. I'm receiving countless messages asking about how expensive it is and if it's working.
So here are those answers:
Your CEO should be strong.
Your CTO should be wise.
Your COO should be wicked, cunning, of mysterious origins, fluent in the dark arts, blurry in pictures,
You will be hearing a lot more about governance and a โcontrol planeโ for AI over the next few quarters.
This is what 8090 built Software Factory for.
Find out more at https://t.co/UAxNYUWgf0
How to make a key employee feel like an owner without giving them ownership:
Phantom equity.
You (and your partners) own 100% of the business. You create a separate document called a phantom stock program. It mirrors your equity but lives outside your operating structure.
It says: โIf we ever sell this business, you get X%.โ
No liability. No tax consequences. No changes to your LLC or corporate docs.
But if they leave, it goes away.
Now your A player is building the business like they own it. Because in the outcome that matters most, they do.
It's the cleanest way to recruit, retain, and align without giving up control.
Polsia just raised $30M at a $250M valuation.
Approaching $10M annual run rate.
One Founder + AI. Zero employees.
Polsia runs companies autonomously.
It also ran its own fundraising.
I just showed up for signatures.
A lot of brilliant AI people back onto the market. If you've been sitting on an AI startup idea @amitpaka โ๏ธ (Founder/COO of @fiddler_ai , $100M raised) and I are running a 2-hour workshop on Maven.
๐ ๐๐ฟ๐ผ๐บ ๐๐ ๐๐ฑ๐ฒ๐ฎ ๐๐ผ ๐๐ป๐๐ฒ๐๐๐ผ๐ฟ-๐ฅ๐ฒ๐ฎ๐ฑ๐ ๐ฃ๐ถ๐๐ฐ๐ต
A 2-hour build workshop on Maven July 14, 10am-12pm PST
๐ฌ๐ผ๐'๐น๐น ๐๐ฎ๐น๐ธ ๐ผ๐๐ ๐๐ถ๐๐ต:
โ A sharpened pitch, pressure-tested live by a former VC and a founder who's raised $100M
โ A specific diagnosis of how to make your story land
โ The frameworks for the five artifacts VCs actually read, positioning thesis, long-form narrative, deck outline, one-page memo, target partner list
โ A 7-day punchlist exactly what to write, in what order, to be investor-ready
โ A take-home Claude skill to help you do the writing well, on your own time
โ Pattern-recognition from watching 14 other founders pitch and get critiqued
Between us: $100M raised at Fiddler, 200+ AI startups evaluated at Storm Ventures.
๐๐๐ฒ๐ฟ๐ ๐ณ๐ผ๐๐ป๐ฑ๐ฒ๐ฟ ๐ฝ๐ถ๐๐ฐ๐ต๐ฒ๐ ๐๐ ๐น๐ถ๐๐ฒ ๐ณ๐ผ๐ฟ ๐ฒ๐ฌ ๐๐ฒ๐ฐ๐ผ๐ป๐ฑ๐ ๐ฎ๐ป๐ฑ ๐ด๐ฒ๐๐ ๐ฟ๐ฒ๐ฐ๐ผ๐ฟ๐ฑ๐ฒ๐ฑ ๐ณ๐ฒ๐ฒ๐ฑ๐ฏ๐ฎ๐ฐ๐ธ. ๐๐ฎ๐ฝ๐ฝ๐ฒ๐ฑ ๐ฎ๐ ๐ญ๐ฑ ๐๐ผ ๐๐ฒ ๐ฐ๐ฎ๐ป ๐ฎ๐ฐ๐ฐ๐ผ๐บ๐บ๐ผ๐ฑ๐ฎ๐๐ฒ ๐ฒ๐๐ฒ๐ฟ๐๐ผ๐ป๐ฒ.
Here's why we created this workshop: I evaluated 200+ AI startups as a VC at Storm Ventures, and the recurring pattern was that smart founders with interesting ideas consistently showed up with pitches that were hard to follow. When the story doesn't hold together, it's hard to build conviction. As a potential champion when I can't quickly understand your company, than I can't advocate for you. And I get founders asking me for this advice all the time.
Link to enroll ๐
https://t.co/BSDzj1hu6O
Wander pretending to be holier than thou. I would sit this one out.
My family trip to a Wander Yosemite tranquil paradise turned out to be a construction zone. Wander support said โdidnโt you read the fine printโ - did nothing. They didnโt let me write a review - knew what was coming. Disingenuous.
[brief update]
we have a new set of funds for @haystackvc
announcing haystack 8 plus more, but it's really about my teammates @aashaysanghvi_ & @DivyaDhulipala
thank you to all of our frequent ecosystem co-conspirators, our LPs, our co-investors, our founders. i'm very grateful ๐ and this is the most fun role in the world!
https://t.co/BwjSh5NXio