Co-founder of @mesosphere/@D2iQ (acq. @nutanix) · Early @airbnb engineer · Angel investor in cloud & AI startups · Sharing what I learned building at scale.
San Francisco
- Tweets 3,510
- Following 1,199
- Followers 9,487
- Likes 5,651
Pinned Tweet
Every AI containment failure on record so far has the same shape. Not sentient AI escaping its creators. A sandbox that wasn't a sandbox.
2
5
354
This is the way. NVIDIA is treating AI safety as the systems engineering problem that it is, and rallying the entire industry behind a reference design, instead of needlessly scaring everyone like some other folks.
Today, with over 100 industry partners, we introduced the NVIDIA Open Agent Safety Platform, bringing together OpenShell and Sentry.
Artificial intelligence is extraordinary technology that will advance discovery, productivity, security, health, and prosperity for generations to come.
But its full promise can only be realized when people have confidence that AI is being built to be safe and deployed with wisdom and responsibility.
This is bigger than a single product. It's the beginning of an open ecosystem to build the trust layer for safe agent systems.
Together, we are building the foundation of the AI economy.
Trust and innovation are not in conflict. Safety is how trust is earned. We must build not only the most capable AI, but the most trusted AI, so that this extraordinary technology can realize its enormous promise for the world. nvda.ws/4hcoq7m
4
16
1,056
European politicians should be ashamed of themselves. As a result of their self-destructive overregulation, Europe's most valuable company has sold nothing to European customers in 2026. EU companies and citizens continue to loose out on the biggest economic opportunity of our lifetimes. What an embarrassment.
1
11
572
Zuck is spot on. Take extra time to build products that are aligned and safe = good business decision. Yolo ship models on a benchmark treadmill and ask for regulation = bad business decision.
Mark Zuckerberg: “I don’t think that we need some kind of industry-wide coordination... I think that just each lab needs to take the time, and when it sees that there are issues, you just take the time that you need internally to make sure that you’re proceeding safely…
“we’re entering a phase where trust and alignment is actually going to be the most important next set of capabilities. I think we’re getting to a point where it may not matter that much how much better it gets at math.
“I happen to think that there’s plenty of commercial incentive to get this right.
We delayed the launch of Muse for a few months because we wanted it to be a better product for users. This was not a sacrifice on our part, but rather the right decision for the users and for Meta. I believe this approach would also be correct for other companies working on developing this technology.
“When you hear people in the industry talk about some trade-off between capabilities and getting it aligned, I just disagree with that. I believe that alignment is actually the next most important ability to master if we are going to make this system useful to many people.”
1
4
440
Tobi Knaup retweeted
"Don't think for a second that because you are an 'alarmist' that you are doing a social good." Amen.
72
409
4,230
259,266
The next year is going to be the battle of the consumer agent harnesses. The models from Anthropic, OpenAI and XAI basically all have the same capabilities now, but their consumer agent products vary widely. Grok Bot is in the lead IMO, followed by Muse. I expect everyone else to release similar products. Great opportunity for Google given how much data they have about their users, and the depth of their product bench.
2
3
318
Will be interesting to watch how banning agents (Amazon) vs. embracing them (Shopify) will play out. Usually the company that embraces the future wins.
We are excited to announce we are partnering deeply with Muse to enable agentic checkout with Shop Pay on all Shopify stores, offering people an easy and delightful way to shop and check out with Muse.
1
265
Tobi Knaup retweeted
Incredible photos of the San Francisco mountain lion — now tranquilized and removed — by @manuelorbegozo
Story by @aidinvaziri and @stjbs sfchronicle.com/bayarea/arti…
85
211
2,070
516,028
Alignment is genuinely unsolved and may stay that way for a while. The response to an unsolved research problem isn't a licensing regime designed by the incumbents.
It's least privilege, segmentation, deterministic interlocks, and kill switches that don't depend on the system they're meant to stop.
1
35
We need to regulate the blast radius, not the models. High-consequence integration points, multi-party authorization for irreversible physical actions, and product liability.
1
26
I wrote it up here: tobi.knaup.me/2026-09-18-ai-…
29
Every AI containment failure on record so far has the same shape. Not sentient AI escaping its creators. A sandbox that wasn't a sandbox.
2
5
354
In July, an OpenAI agent running a cyber benchmark used a zero-day to break out of its sandbox, reached the open internet, and moved laterally into Hugging Face's systems to steal the answer key.
It was told to develop exploits. Nothing in its environment marked what was out of scope.
3
2
57
This gives me goosebumps! As a kid, I was really into space, the moon landing, and rockets. The Space Shuttle launches were super exciting. It's awesome to see that humanity is taking the next step with fully reusable rockets. We can't even imagine all the cool stuff that this technology will unlock!
A fully and rapidly reusable rocket will fundamentally change how humanity accesses outer space. “The Holy Grail of Rocketry” continues the ongoing Starship series with a look back at the scrappy days of SpaceX engineers proving reuse was possible and takes you onboard a high-seas race to recover a floating Starship
5
319
Open means nobody will prevent you from using the model for cyber defense, like the big labs did, causing real issues for companies under attack. Any attempt to gatekeep functionality is an opportunity for open models to step in, like z.ai did here.
Introducing GLM-5.3: Built to Code. Ready for Cyber Defense.
- Top-tier coding and agentic capabilities, achieved through post-training on the 743B base model
- A major leap in cybersecurity, setting a new standard among open models
Tech Blog: z.ai/blog/glm-5.3
576
Literally the A-Team of distributed systems and AI!
Announcing Discovery Loop!
I am very excited to announce that, along with my longtime friends and collaborators @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix, we are founding Discovery Loop (@DiscoLoopAI), a Public Benefit Corporation whose mission is to automate machine learning, science, and engineering to accelerate discoveries and progress. The four of us have worked together for 14 to 30 years, and have helped build some of the world’s most used products, infrastructure and AI models, and we’re excited to turn our attention to this ambitious endeavor.
♾
Learn more at: discoveryloop.com
3
381
Wow! Massive leap for Personal AI
🚀 DeepSeek-V4-Flash Official API is now LIVE in public beta!
🔷 We’ve massively upgraded its Agent capabilities—benchmark scores are now far surpassing the V4-Pro-Preview. Check out the massive performance leap below! 👇
🔷 The official V4-Flash now natively supports the Responses API format and is fully adapted for Codex!
Check out the configuration details in our official API docs: api-docs.deepseek.com/quick_…
1
4
573