Born Again Enterprise | 🇵🇭 🐂 UT CS '15 | Empathy Before Profits | dstack
Warlox
- Tweets 6,737
- Following 873
- Followers 2,923
- Likes 19,929
Joshua Waller retweeted
Venice AI runs its private TEE models on Phala.
Prompts are processed inside GPU TEEs, and every run is hardware-verified.
$PHA
3
19
104
4,873
Phala should be $1B mcap at the least.
Not a better engineering team in Confidential Computing and AI out there.
Made Private AI end to end possible for every major project out there through dstack.
True innovators in OSS that are humble and share their knowledge.
6
5
42
1,087
Dependencies that change ways of working take 12-18 months to enter a large enterprise at scale. How do you reduce that timeframe?
You can't. Gotta play into the strategy & understand how your product changes as little as possible to get integrated at scale.
196
This is quite hilarious 😂 Every industry leader has the same memo idea. Sidekicks/digital twin/shadow agents, etc.
Muse is unique for 1 reason; Network Effect. I'll be impressed if users stay active after 3-6 months though then that means you have executed on what others have yet to deliver on
if you wanna know how long we've been working on muse, here's an excerpt from a doc that i wrote for the meta board a year ago in sep 2025
@natfriedman, the whole team & i have been building muse for a long ass time. been a real pleasure to bring that to life for all of you ♥️
2
1
328
Joshua Waller retweeted
Yes we are a healthcare company. Yes we just released the best open-weight alternative to Jev, Solomon 27b.
- It roughly matches the intelligence of its base-model, Qwen3.8 27b
- Natively Multimodal
- 265k Context Window
- Adds multi-choice tagging, evidence pointer spans, and more.
37
55
733
56,508
Joshua Waller retweeted
Phala Confidential AI: 254.51 billion tokens over the past 24 hours.
Top models:
GLM-5.3 70.49%
Kimi K3 10.73%
GLM-5.3 Flash 8.18%
phala.com/models/z-ai/glm-5.…
TEE-backed AI infrastructure.
12
20
127
8,138
If I had Jev last year in January, I would have a fully managed NixOS setup end to end 🤣
I had looked into training a small model to handle instructions, but my RL skills are trash 🫠 so this project sat in the corner of innovative products with the label "skill issue".
I told you all, I can tell the future. I have the best instincts for these types of products based on my experience.
NixOS management agent will be the next one thats too early 🥲
Imagine the power of an OS you can seamlessly upgrade and rollback when the agent makes mistakes. Makes me sad to reflect on this
1
332
Quick experiment with @typesafeai Jev as a semantic-analysis primitive inside a deterministic product-delivery compiler. Like I keep mentioning, enable automation with more certainty
I don't want the model to choose the workflow, skills, concerns, or next stage. Let code own those decisions. Use Jev only to answer narrow semantic questions about a value streams desired outcome then let humans accept that risk w/ recorded evidence
2
1
248
Lmao you actually dont need skills. You just need structure to deterministic repeatable actions and resources.
Eliza was missing Jev.
1
149
Blah blah blah MCPs wins.
Well of course they win. What self respecting enterprise company will allow autonomous agents to run bash commands with no guardrails? Granted, building a governed capability broker you're fine but security teams wont allow agents to rift in the CLI
161
At Phala, we solved so many HARD problems. Problems you only find in the deepest parts of the ocean. I would not be as great as I am today without the mentorship & hardwork done during the last 4 years.
I miss the work everyday, but they prepared me for the weight I carry today
1
12
586
When will Human Giant be on streaming services so I can show my wife & kids the realities of having fun outside & the cutthroat world of child acting? @azizansari @robhuebel @paulscheer
101
I'm forcing my team to implement a rules filter that is deterministic when dynamically building context. They want to use an agent or knowledge graph, but why introduce less certainty?
2
3
398
Joshua Waller retweeted
yeah, we don't need new laws. we just need existing laws to apply. i.e. if my software company hacks another company that would like get me jailed. doesn't matter who or "what" did the hackeroo.
Jensen Huang pushed back hard against AI labs calling for tighter regulation and slower AI development, at Dreamforce.
"If you are not confident about the safety of the products, and you're not confident in its functionality, capability, or safety, then don't just release it.
We don't need any new laws. We don't need new regulations. we just need companies to decide that when it's time to run as fast as they can."
safety and speed are not mutually exclusive, and new laws are not required to make AI systems safer. Build fast, test aggressively, and pause only when you are no longer confident the product is safe.
---
From "Salesforce" YouTube channel, (full video link in comment)
12
34
425
33,848
My product being used across the firm to demo to clients. It only gets better everyday. Fun part was building the brand kit to transform the interface to the design system of the client they're pitching to. Fun times.
2
1
275
I remember seeing this pitch deck awhile back. Palantir does a great job on their pitches. I see some firms try to reimplement the model, but the work that goes into building the ontology is not easy then adding compliance and standards is another fun problem to resolve in enterprise environments. Those projects live & die based on the FDEs at helm
What operating a Sovereign AI stack looks like
@nvidia + @PalantirTech = 🇺🇸
1
257
Genius. Team is always delivering simple 🫡 glad they chose business side.
Rewrite rules for managing the input context with scoped access is what all businesses want. If I understand correctly, Context Control Lists for Model Routing makes impact like ACLs for Network Routing
If you want to play around with an early thing: you can now give us money and we give you tokens in return. radius.earendil.com/
1
1
475
