Usman

Building something new for physical R&D. Previously founded Amazon Bedrock and AWS #IIoT. @yale engineer

[Paul Graham @paulg]

One way you can recognize AI slop is when the ideas and the diction don't match — when the ideas are completely ordinary, but the diction is that of someone announcing a brilliant discovery that they're really excited about.

[Usman @uanwer]

Enterprise service providers aren’t and can’t be divas. Provide a great service and reduce cognitive burdens of your customers. You shouldn’t be a genius the customer has to learn to deal with.

[Corey Quinn @QuinnyPig]

Remember: being a co-founder means `main` is your feature branch.

[Usman @uanwer]

Seattle is for the long game and civilizational infrastructure from trucks to jets to the PC to e-commerce to the Cloud and also coffee! Going on to Satellite constellations and nuclear fusion. Not great for get rich quick schemes but that’s the point.

[Aidan T🌐 @buildhomez]

It's crazy that Seattle is guaranteed to dominate the next decade more than the last one.

The cloud capital of the world during a data center boom.

The space telecom capital of the world as SpaceX nears $3T cap.

Offices of all major AI labs as SF housing languishes.

[Usman @uanwer]

And even on commodity work, cost + control isn't an automatic win. What closed still has going for it:

benchmarks aren't private evals, so close doesn't mean conversion
cheap closed options already exist (Sonnet, GPT minis)
serving quality: capacity, latency, throttling
a neck to choke when it breaks
commit discounts that beat list price
self-hosting carries its own cost of ownership

Open wins where good-enough + control clears all of that. That set is growing for sure, but it isn't everything.

[Usman @uanwer]

The open vs. closed model debate is mostly noise because it skips the only question that matters: is the workload commodity or frontier-bound? Frontier work stays closed. Commodity work is already in play. The ships don't turn all at once. They turn workload by workload.

[Kartikeya Tanna @KartikeyaTanna]

Ironically, the world’s first trillionaire held an H-1B visa.

[Usman @uanwer]

This seems less about AI regulation and more about resolving concerns between key partners effectively. I hope the lesson from this is more listening on behalf of customers, who were the ones impacted at the end and expect stability from enterprise service providers. It’ll be a while before the customers forget it though. Multi-model will be even more important to all enterprises.

[David Sacks @DavidSacks]

I’ve had a number of conversations with folks inside and outside government about the current situation with Anthropic, and here is what I believe to be true:

— As we know, Anthropic publicly released its Mythos class models earlier this week under the commercial name Fable.

— Fable is Mythos with guardrails. But if those guardrails fail, then you’ve exposed Mythos and its advanced cyber capabilities to people who shouldn’t have them. (Keep in mind that Anthropic itself widely promoted the idea that Mythos was a cyberweapon and needed to be regulated as such. They asked for government regulation of Mythos and championed the guardrails on Fable. If there is a vulnerability — big or small — it is Anthropic’s responsibility to patch.)

— A highly credible trusted partner of both Anthropic and the USG who was testing Fable came forward with a jailbreak of those guardrails. The Admin asked Dario to fix the jailbreak or de-deploy the model. Dario refused.

— In their blog post, Anthropic defended its decision by saying the jailbreak isn’t serious. That is not what the trusted partner and the USG believe; nor is that kind of minimizing language consistent with Anthropic’s brand as the AI safety company. It’s difficult to fathom how they could claim a jailbreak allowing operability of a cyber weapon could be defined as not “serious.”

— In the past, Anthropic has always said that safety must be top priority and taken super seriously. In this case, Anthropic prioritized the continued offering of the consumer model over safety.

— In reaction, the Admin issued the export control. The Admin did this reluctantly. It’s been very surprised that Anthropic hasn’t wanted to cooperate with a reasonable safety request (ie fixing the jailbreak issue). Anthropic’s reaction is very much at odds with their branding and ethos as a safe AI research community.

— The Admin’s hope now is that Anthropic remediates the safety issue, the export control is lifted, and Fable goes back into general release. The Admin wants all of this to happen as soon as possible. It is frankly bewildered that Anthropic hasn’t wanted to comply with safety requests that it previously said were its highest priority.

— Those trying to misdirect and tie this action to the prior DoW/Anthropic issues are wrong. The Admin values Anthropic’s technical capabilities and feels that this issue, while serious, should be easily resolved. The ball is in Anthropic’s court.

[Usman @uanwer]

In my ten plus years of building enterprise platforms I couldn’t imagine a scenario where a core tech provider has to turn off that capability on a moments notice. Would AWS be AWS and Microsoft be Microsoft if they did that - like ever? You can’t run a business like that. It’s absolutely unacceptable. Any tech that ends up that way simply isn’t enterprise ready.

[Usman @uanwer]

This! Not hiring new grads is a strategic mistake and frankly an indication of a second rate leadership. You get more ideas and perspectives out of this cohort than any existing one at your company.

[Scott Kennedy ⠕ @stkenned]

One of the worst takes in tech right now is not hiring new grads.

At Replit, we're arbitraging an industry mistake.

We doubled our new grad hiring targets and they're onboarding now.

(full post in 🧵)

[Usman @uanwer]

Agree with this. Anthropic is doing the right thing, maybe the messaging could be better. We actually can’t use the model for anything since we touch physical R&D, but precisely because of that we know that unrestrained frontier AI can be weaponized pretty quickly. Better to be annoyed than unsafe.

[Seth Bannon @sethbannon]

The backlash to Anthropic's limited Mythos/Fable rollout is bizarre.

For years, an argument against AI regulation has been: "Trust companies to do the right thing."

Anthropic makes a decision that costs it money to ensure safety and suddenly that's unacceptable?

[Usman @uanwer]

Paid a top freelancer for an animated product explainer. Got it this morning. Then made one with Fable from scratch in a single evening with new ideas included, down to a T. It is the best of times. It is the worst of times.

[Usman @uanwer]

Great to see Harvey's legal agent benchmark included in the Fable model card and the shift towards real workflow benchmarks outside of coding. The numbers show that there's so much to left to do.

[Usman @uanwer]

Jokes aside, the model is more helpful. Sharper nuance on the knowledge tasks I use Claude for daily as an entrepreneur. Echoes my early Opus 4.6 experience, but more palpable. Nice work, team @Anthropic.

[Usman @uanwer]

We are so cooked!

[Usman @uanwer]

What we all really need is a Claude phone.

[Usman @uanwer]

Finally vibecoded myself the most important app - figures out that I may have fallen behind and reminds me to connect meaningfully with parents.

[Usman @uanwer]

Always refreshing to read a contrarian perspective. If this is about innovation/collaboration and not labor costs, we could also ask the chinese companies getting the billions to open US labs and hire US talent. Also we must ensure we get all discovery data from offshored deals for model training.

[nature @Nature]

A single dose of engineered immune cells has helped two men and one woman to receive life-saving kidney transplants. Their bodies would normally reject donated organs.

https://t.co/CyiKUBpuPc