vansh taneja retweeted
Can we forecast benchmark accuracy from models that score zero?
We (co-leads @LukeBailey181 @rckpudi) turn a benchmark into a series of easier hinted versions. Jointly modeling scale C and hint level H enables forecasting GSM8K acc (H=0) within 3 pts from 5600x less compute 🧵
vansh taneja retweeted
Summary of CPO/NPO Market Update
🚀 NVIDIA CPO: Spectrum-X CPO switches for Scale-out are officially in mass production. Our forecasts: 15k units (2026E) & 100k units (2027E).
*Supply Chain: To support the CPO Scale-Out ramp, TSMC has expanded CPO inspection equipment capacity (notably adding Insertion 2/3 capacity), alongside strong progress from key suppliers in FAU, shuffle boxes, and system assembly.
* Scale-Up Architecture: Rubin Ultra is now expected to adopt a 9-18-9 tray design. While mechanical challenges from the 0.75U height could potentially lead to a reversion to a 10-9-8. This shift will have no impact on optical engine (OE).
📈 OE Shipments: NVIDIA platform optical engine (OE) shipments expected at 6m (2027E) & 19m (2028E). Total industry-wide OE hitting 11m & 40m.
☁️ Amazon Trainium 4: AWS projected to consume 5m (2H27E) & 12m (2028E) OE units, primarily 6.4T specs.
* Trainium 4 will likely have 3 configurations, with two expected to adopt NPO.
💡 Key Plays: LITE/COHR (CW laser upside), Browave (10k+ shuffle box per quarter in 4Q26E, followed by acceleration in 1H27E.), SMTC/MRVL (TIA/Driver), TSEM (NPO PIC exposure).
#NVDA #LITE #MRVL #SMTC #TSEM #Browave #CPO #NPO
vansh taneja retweeted
I spent an hour solving the fourth IMO problem last weekend. Then I checked Pramaana's performance - which formalized and solved it in half the time, machine-checkably correct. Not sure if I should be a proud founder or a sad mathematician 🥲
Replying to @PramaanaLabs
@PramaanaLabs prover, Hardy, took a quick break from verifying mission-critical reasoning across tax, legal, and financial domains last weekend to tackle something a bit more recreational: IMO 2026.
Hardy solved 6/6 problems in 7 hours using Lean 4. Our autoformalization model, Panini, translated the natural language problem statements into Lean 4.
Full write-up & GitHub repo with Lean proofs: pramaanalabs.ai/blog/pramaan…
vansh taneja retweeted
Today, I'm thrilled to announce Pramaana's $27M seed, led by @khoslaventures.
The foundational domains that hold the world together: tax, law, finance, healthcare; all run on certainty. Probabilistic AI can't give them that. We’ve been asked to accept wrong answers with AI as ‘hallucinations’, while in traditional software terms, it’s just a bug. And a wrong answer in such mission-critical domains is more than just a bug, it's a liability that could have catastrophic impact.
We built Pramaana to deliver a 100% trustable experience to the domains that run on certainty: AI that is provably correct, not probabilistically correct. We turn statute and regulation into machine-verifiable code, so every output ships with mathematical proof of correctness. Our mission is to make AI take ownership of it’s work.
Pramaana in Sanskrit stands for “means of valid knowledge”, and we’re going to achieve that by formalizing the world’s knowledge.
vansh taneja retweeted
watch Owl Posting podcasts to be aware of the future roughly 8 months in advance
owlposting.com/p/can-machine…
Together with UC Berkeley we are announcing the laser phase plate - a breakthrough in atomic resolution imaging. This is the brightest continuous wave laser in the world, 100 million times the intensity of the surface of the sun.
Phase contrast plays an important role in microscopy, but it was thought close to impossible for electron microscopy, where it would require interfering with an electron beam. Holger Mueller and Robert Glaeser proposed exactly this using a standing wave laser. It has taken over 15 years to make this a reality. Biohub partnered with UC Berkeley and Mueller to support this work and to engineer and build the technology.
Contrast has been the critical barrier to achieving atomic resolution imaging of the cell. In cryo-electron tomography, a cellular imaging technology that uses electron microscopy, the low contrast makes it impossible to resolve anything but the largest proteins within their cellular context. The laser phase plate removes that barrier.
With advances in AI this breakthrough in contrast will start to open up a new frontier in structural biology, that will allow us to see the molecular machines of the cell, and how they assemble into far more complex and dynamic systems, and understand how they work.
vansh taneja retweeted
Glad this is (finally) being talked about. let me walk you through the issues in chatGPT memory and how @supermemory solves them.
1. lack of forgetfulness and lack of temporal understanding
Notice how chatgpt saved "wants to travel SF today"
that was 3 months ago.(more below)
vansh taneja retweeted
All-time favorite type of videos: pre-fame bands playing their extremely famous songs to a tiny room of people, because they're not yet known. A Thread 🧵
1. Two dudes in 2003, unaware they were making a legendary song
vansh taneja retweeted
Most startups fail because they chase scale before truth. The product isn’t loved. The pain isn’t real. But the pitch is polished. You can raise on narrative. You can’t build on fiction.
vansh taneja retweeted
Cursor round closed — $625M at $9.6B post led by Thrive & A16z. Accel is a new backer.
$200M ARR, up 4x from $2.5B round in November 2024.
ARR multiple constant from last round at 50x.
In less than a year, the startup behind Cursor has sparked tech's buzziest buzzphrase—vibe coding—and seen its valuation go from $400M to $2.5B to possibly $10B.
It's one of the fastest-growing startups ever. @nmasc_ has the inside story for Weekend. theinformation.com/articles/…
vansh taneja retweeted
The biggest bottleneck isn’t execution speed. It’s the lag between action and feedback. If you optimize for speed to feedback, everything else gets easier.
vansh taneja retweeted
If you want to accelerate growth, fix your feedback loops. Track how long it takes to go from action to insight. Shrink that number and you’ll shrink the time it takes to get results.
vansh taneja retweeted
Manus just blessed me AGAIN
They saw how quickly we ran through the first 100 community codes… so they gave me 500 more….
If you missed out on the first round, here’s your chance:
1. Retweet this post
2. Follow
3. DM me “MANUS”
If you got the access code in the first round, reply with what you cooked up so far!
Manus is absolutely next level. I literally maxed out the tokens yesterday.
I was blessed by the AI gods and have exclusive beta access to Manus…
Here’s my key takeaways:
-it’s incredibly powerful, you can run up to 50 browsers at once
-it’s like combining operator + clause computer + deep research
-I haven’t had to step in once to override, and it’s running a task for 2 hrs straight
their waitlist is now almost at 1 million people…
but I got insider access for my Twitter community…
first 100 people to repost and follow will get my access code to try Manus yourself!
vansh taneja retweeted
Today we're releasing the #1 interface for MCP Servers. Everything you already love about Cursor, across every app on your computer.
What Apple Intelligence & Copilot could have been.
1) Screen grounding: @ any service in-context, and write just by hitting tab.
2) Scroll 👇
vansh taneja retweeted
The barrier to entry for building anything is dropping to near zero because of AI. Designers can build bring their apps to life. PMs can turn their PRDs into prototypes. IT teams can build custom software for anything. Software is about to go through a huge growth spurt.
vansh taneja retweeted
I see this over, and over, and over again. Especially in relation to AI:
"Never expect someone to understand something that their paycheck depends on them not understanding."
It's too psychologically painful to believe that the job you spent decades on could become irrelevant.
vansh taneja retweeted
One of the biggest mistakes I see people make is they don’t start a company because they don’t have that “perfect” idea.
An idea will never be perfect, it will always evolve but activity drives learning and iteration.
Start, observe, iterate and move fast.



































