Vikas Shanbhogue

[Arya Hezarkhani @_i_am_arya]

forward deploying the frontier. ruggedized. offline.

R2D2 goes live in 2026.

powered by @AMD Strix HALO.

[Nico Bykhovsky @BykhovskyNico]

Hardware has historically suffered from catastrophic TTDH (time to dopamine hit).

Atlas Desktop controlling a HackRF Pro over USB in one shot.

We’re on the best possible timeline @GSGlabs

[Trevor B (Trev The Creator) 🇬🇾🇺🇸 @Trevs_Dev]

Very proud of the craftsmanship and precision my team at @arenaPhysica put into this article

Every component of the 2019 DJI Mavic Mini and 2023 Mini 4 Pro, torn down and priced at its floor cost

We've also added an AR experience that shows the teardown in AR

https://t.co/QYxpjBYyYd

#buildinpublic

[Arena Physica @arenaphysica]

Can the US build a DJI? We bought a 2019 Mavic Mini and a 2023 Mini 4 Pro, tore them down in our hardware lab, and put them under a bench microscope to find out.

What we found: https://t.co/2nxcHd8cSF.

[Pratap Ranade @PratapRanade]

Can the US build a DJI?

We tore down @DJIGlobal 's 2019 and 2023 Mini drones and priced every part. It's the first public cost breakdown of either. Here's what we learned:

DJI earns SaaS-like margins on hardware: ~79% gross margin on a $139 parts BOM. For comparison, Apple makes 53% on the iPhone. We estimate US-sourced parts would cost 2-3x more, cutting margins to 37-58%.

The result: US drone makers can't compete in consumer, and focus on premium enterprise and defense.

Sovereign silicon: DJI is building its autonomy lead starting at the chip. In 2019, the drone's key chips were American and Dutch. By 2023, all four were DJI's own.

Custom silicon jumped from 1% to 26% of the parts bill. Nobody else can buy these chips.

One notable exception is the GPS chip, still from Swiss firm @ublox . Chinese alternatives exist, but DJI's flight software is built around u-blox. In modern hardware, lock-in lives in the software.

"Cheap China" is an outdated perception. DJI paid more for lighter Chinese batteries, then spent the saved grams on sensors and compute. Their first priority is capability per gram, not cost.

Better battery tech → lighter weight → more onboard autonomy. US-made @SkydioHQ has better autonomy, but needs a heavier drone and a $399 chip to get it. We lost the war on cost. The next battlefield is autonomy.

DJI's real moat is speed. Shenzhen gets new circuit boards in hours, the US in weeks. "Move fast and break things" works for hardware in China, not America.

Banning DJI doesn't close that gap. Faster iteration does. We ran this teardown with our FDEs and Atlas, our engineering agent platform.

See our full analysis: https://t.co/XHTL8X5uyW

Thanks to @JonathonPSine @AkhilIyer1, Maggie Gray and Shaun Donovan for reading drafts of this.

[Pratap Ranade @PratapRanade]

Today, we’re excited to unveil Heaviside-1, @arenaphysica's second generation foundation model for electromagnetism. 5 months since the release of Heaviside-0, Heaviside-1 is a crucial milestone on our quest to build EM superintelligence: a foundation model that understands EM across the spectrum, from RF through photonics, capable of designing the next generation of electronics.

Major updates:

- Heaviside-1 is >10x the size of Heaviside-0 (roughly the size of GPT-2), trained on 250k unique designs with >500B unique EM field samples.

- It runs 10^5 x faster than commercial solvers, with accuracy <1 dB.

- Heaviside-1 natively encodes 3D structures, their material properties, and excitation patterns as input.
It predicts full EM fields at arbitrary locations in space, not just downstream quantities like S-parameters.

- Most importantly, our bet on fields lets Heaviside generalize OOD proving that it’s learning real EM physics: Heaviside-1 generalizes outside of distribution very well, leaping from 0.99 dB to 0.53 dB S-parameter error.

Try it out in Atlas Fields Studio in beta today (link in replies). I personally had a lot of fun seeing how EM fields twist and curl around different circuits. I wish I had this when I was learning EM in college. I hope you enjoy it as much as we did.

[Arya Hezarkhani @_i_am_arya]

Today, we’re proud to unveil Heaviside-1, @arenaphysica's second generation foundation model for electromagnetism.

Five months since announcing Heaviside-0, today’s launch marks a critical milestone towards building electromagnetic superintelligence, a foundational intelligence that understands EM across the spectrum and is capable of designing next gen electromagnetically-governed systems, from RF to semiconductors to photonics.

Why we’re so excited about Heaviside-1, and what today’s announcement represents:

> It’s big: Heaviside-1 is a GPT-2-size model, roughly 10x the size of Heaviside-0 and trained on 250k unique designs with >500B unique EM field samples
> It’s fast: It generates predictions with 5 orders of magnitude acceleration relative to commercial solvers, with under 1dB of accuracy
> Full-3D Encoding: Heaviside-1 natively encodes 3D structures, their material properties, and excitation patterns as input, a major step towards generalizability across design spaces
> Understands the underlying physics primitive: Heaviside-1 generates full-3D near-fields at arbitrary locations, not just downstream quantities like s-parameters
> Generalization: Most importantly, our bet on fields as the primitive for Heaviside-1 generalizes OOD, proving that it’s learning real EM physics. Heaviside-1 generalizes outside of distribution incredibly well and better than Heaviside-0, down to 0.53 dB s-parameter error from 0.99 dB on Heaviside-0.

Read more about Heaviside-1 and the work that went into it in this write-up by our research team, @m__frei, @chris_m_bryant, @nathanmirman, @HaoLiu69889370, Tommaso, Roberto, Ruichen, Noah, Boyuan, and @Trevs_Dev, and our RF team @hk2532_harish, @anrfic, @suyangcn and Zhaoji.
https://t.co/XM2iwaWb5Z

[Arya Hezarkhani @_i_am_arya]

No matter the day of the week, we’re open for business.

This morning’s focuses: working on reasoning, evals for various new task types, and updates to our training pipeline.

[Arya Hezarkhani @_i_am_arya]

Today, we're introducing @arenaphysica, the electromagnetic superintelligence company.

Our mission: accelerate the design, development, and deployment of the devices that will sense, communicate, compute, and actuate in the future of autonomy, with EMSI.

For years, we’ve been building a new type of company in the shadows with our partners and collaborators.

We're proud to share that we’ve deploying EMSI with @AMD in AI compute, @anduriltech in defense technology, and @SiversSemicond in satellite and 5G communications.

We've reached the Electromagnetic Inflection.

software-defined hardware == electromagnetically-governed device

Their behavior more attributable to Maxwell's equations than Newton's. A car's most expensive subsystem is its electronics. A datacenter's binding constraint is power delivery and SerDes interconnects. The F-35's avionics cost more than its engine.

Electromagnetic design is a physics problem, and at @ArenaPhysica, we're building the stack to solve it.

-> Atlas: an agentic platform deployed as the intelligence system of record for hardware workflows at companies building frontier devices, making probabilistic reasoning engines capable of deterministic engineering work.
-> The EM Foundation Model: the first physics foundation model for electromagnetism, learning the relationship between geometry and fields directly to drive the time and cost of verification toward zero.
-> The Electromagnetic IP Factory: vertical co-development of the most advanced electromagnetic IP in defense, automotive, compute, and communications alongside our most ambitious partners.

I’m proud to be building this alongside @PratapRanade , @_clairepan , @m__frei , @hk2532_harish , and the rest of our artist colony here at @arenaphysica.

Read our thesis: https://t.co/ujKJbWRy9y

And a big thank you to @packyM for telling our story and how electromagnetism (secretly) runs the world.

[Packy McCormick @packyM]

The future is electromagnetic.

One challenge is that there are ~ten people in the world who can deeply intuit electromagnetism. RF engineering is "black magic."

Arena Physica thinks machines can intuit EM better.

CEO Pratap Ranade &amp; I on AI for EM:

https://t.co/U3CUnqcYPp

[Arena Physica @arenaphysica]

Introducing Arena Physica.

For years, we've been quietly building alongside the teams defining the frontier of hardware — embedding with @AMD in AI compute, @anduriltech in defense technology, @BauschLomb in advanced manufacturing, and @SiversSemicond in satellite and 5G communications.

Atlas, our platform for agentic hardware workflows, is only the beginning. We believe a new class of foundation model will let humans push further into the physics that shapes our world but remains fundamentally unintuitive to us.

Read our thesis here: https://t.co/tmRxPcWrRi.

Thank you to @packyM for capturing why electromagnetism secretly runs the world - and what it means to build electromagnetic superintelligence.

Our foundation model launches in 7 days. Link in the comments to be first to know.

[Vikas Shanbhogue @svikas20]

@Ask_Spectrum The DM response of "call our customer service" when I've mentioned that I've done both that and store visits multiple times isn't helpful at all. I would cancel my service, but I can't even do that because nobody seems to be able to find my account (yet somehow I am still billed)

[Vikas Shanbhogue @svikas20]

@Ask_Spectrum My Spectrum Mobile account has been "pending migration" (according to many customer service calls and store visits) for months. I can't see bills, change CC or anything else. Of course I'm still getting billed every month. Nobody will escalate this. What's going on?