Building AI heart for Robots | Autonomous Security Workflows AI Cameras → AI Agents → Humanoid Robots → On-Chain Verification
Building the Future
Joined June 2026
- Tweets125
- Following30
- Followers1.4K
- Likes257
Pinned Tweet
Most "AI agents" are text in a box.
This one has legs.
Me @sar1hakkk Meet my friend SecureRob 🤖 — an autonomous security workforce that actually walks up to you.
Real humanoid. Real patrol. Fully on-chain.
Here's a look 🧵
AI is learning to do more than think.
From executing tasks to interacting with humans, physical AI is bringing intelligence into the real world.
At SecureRob, we’re building toward that future one step at a time. 🦾
#PhysicalAI #Robotics #EmbodiedAI
This Week in SecureRob: Building the Infrastructure for Physical AI
This week was about going deeper than the robot itself.
We’re building toward a future where robots don’t operate as isolated machines, but as intelligent agents that can perceive, reason, act, and remain accountable in the physical world.
Here’s what we’ve been working on:
🟩 PHYSICAL AI
We’ve been developing the architecture for robots to operate as part of a larger sensed environment rather than relying entirely on their own onboard perception.
The core idea:
The building senses. The robot subscribes.
🟩 WIFI SENSING
We’ve been exploring building-side RF/WiFi sensing using Channel State Information (CSI) to understand presence, movement and occupancy without relying entirely on cameras.
The goal isn’t to turn WiFi into another surveillance layer.
🟩 ROBOT + ENVIRONMENT
One of the biggest architectural shifts we’re exploring is treating perception as a property of the space, rather than something every individual robot has to independently recreate.
🟩 ACCOUNTABILITY
Physical AI needs more than intelligence.
It needs evidence.
We’ve been working through an architecture where robot decisions can be cross-referenced against independent environmental observations, creating a stronger record of what actually happened.
The direction we’re exploring is:
Sensing → Belief → Policy → Evidence → Action
🟩 SAFETY FIRST
One principle we’re keeping hard-coded into the architecture:
Uncertain sensing should only make a robot more cautious — never less.
If an RF signal suggests a corridor might be empty, that shouldn’t make a robot move faster. If there’s uncertainty, the system should slow down, investigate or alert.
🟩 THE CALIBRATION FLYWHEEL
We’re also exploring how robots can help solve one of the biggest problems with WiFi sensing: domain shift.
A robot already knows where it is and can build a calibrated map of its environment. Its normal operation can therefore generate valuable pose-to-CSI data and help continuously calibrate the sensing layer around it. physical-ai-wifi-compliance-x-article.md
🟩 WHAT’S NEXT
We’re starting with the fundamentals:
→ Ambient presence
→ Robot attention routing
→ Environmental calibration
→ Cross-system evidence
→ Stronger policy enforcement
→ Carefully validated higher-level applications
The long-term vision is simple:
Robots that can perceive beyond their own line of sight.
Buildings that can understand what’s happening inside them.
And an evidence layer that can prove what happened.
That’s the direction we’re building toward at SecureRob.
Physical AI isn’t just about making robots smarter.
It’s about making the entire environment around them intelligent, accountable and trustworthy. 🤖🦾
We’re just getting started.
We’ve spent years figuring out how to make AI manage money.
ACP, EconomyOS, Butler on X, agentic trading competitions. We’ve tried different approaches, learned what worked and had to rethink what didn’t.
Personal AI services keep getting smarter. Getting them to put your money to work is the next big step.
Introducing Virtuals App, now in closed beta.
Your personal AI, with a wallet and access to markets. Talk through ideas, research investments and put money to work, all inside your group chat with friends.
SecureRob retweeted
We’ve spent years figuring out how to make AI manage money.
ACP, EconomyOS, Butler on X, agentic trading competitions. We’ve tried different approaches, learned what worked and had to rethink what didn’t.
Personal AI services keep getting smarter. Getting them to put your money to work is the next big step.
Introducing Virtuals App, now in closed beta.
Your personal AI, with a wallet and access to markets. Talk through ideas, research investments and put money to work, all inside your group chat with friends.
Physical AI is the future!
Moving AI into the physical world makes sense, but obsessing over the humanoid form factor just wastes the actual opportunities of robotics by needlessly reproducing human limitations like two legs and arms.
#Robotics #AI
xitter.cf/secure_rob/status/2106…
SecureRob retweeted
Why did we build SecureRob?
We believe the next chapter of AI won’t happen inside a chat window. It will happen in the physical world.
For years, we’ve watched AI become increasingly capable of understanding, reasoning, writing, and making decisions. But there’s one frontier that still has so much room to explore:
Giving AI a body.
That’s the idea behind SecureRob.
We started with a simple question:
What happens when intelligent AI agents can interact with the real world through robots?
Not just responding to prompts, but observing their surroundings, understanding their environment, learning movements, performing tasks, and continuously improving through real-world experience.
But we also believe intelligence alone isn’t enough.
As AI moves into the physical world, reliability, security, and trust become just as important as intelligence. A robot operating in the real world needs to be dependable, safe, and capable of handling situations beyond controlled environments.
That’s the future we want to help build.
What’s ahead for SecureRob?
→ Bringing AI agents closer to real-world robotic hardware.
→ Exploring embodied AI and autonomous robotic systems.
→ Improving how agents perceive, interact with, and learn from their environments.
→ Building toward more reliable and secure human-robot interactions.
→ Contributing to an ecosystem where intelligence doesn’t stop at software.
We’re still early. There’s a lot to experiment with, a lot to learn, and an incredible amount of engineering ahead.
But every experiment, every movement, and every breakthrough brings us one step closer.
We don’t just want AI to think about the world. We want AI to become part of it.
This is SecureRob. The journey from digital intelligence to physical intelligence starts here. 🤖🦾
#SecureRob #PhysicalAI #EmbodiedAI #Robotics #Virtuals
SecureRob retweeted
Most "AI agents" are text in a box.
This one has legs.
Me @sar1hakkk Meet my friend SecureRob 🤖 — an autonomous security workforce that actually walks up to you.
Real humanoid. Real patrol. Fully on-chain.
Here's a look 🧵
Why did we build SecureRob?
We believe the next chapter of AI won’t happen inside a chat window. It will happen in the physical world.
For years, we’ve watched AI become increasingly capable of understanding, reasoning, writing, and making decisions. But there’s one frontier that still has so much room to explore:
Giving AI a body.
That’s the idea behind SecureRob.
We started with a simple question:
What happens when intelligent AI agents can interact with the real world through robots?
Not just responding to prompts, but observing their surroundings, understanding their environment, learning movements, performing tasks, and continuously improving through real-world experience.
But we also believe intelligence alone isn’t enough.
As AI moves into the physical world, reliability, security, and trust become just as important as intelligence. A robot operating in the real world needs to be dependable, safe, and capable of handling situations beyond controlled environments.
That’s the future we want to help build.
What’s ahead for SecureRob?
→ Bringing AI agents closer to real-world robotic hardware.
→ Exploring embodied AI and autonomous robotic systems.
→ Improving how agents perceive, interact with, and learn from their environments.
→ Building toward more reliable and secure human-robot interactions.
→ Contributing to an ecosystem where intelligence doesn’t stop at software.
We’re still early. There’s a lot to experiment with, a lot to learn, and an incredible amount of engineering ahead.
But every experiment, every movement, and every breakthrough brings us one step closer.
We don’t just want AI to think about the world. We want AI to become part of it.
This is SecureRob. The journey from digital intelligence to physical intelligence starts here. 🤖🦾
#SecureRob #PhysicalAI #EmbodiedAI #Robotics #Virtuals
All eyes on @secure_rob !👀
Catching up on Virtuals: the latest across the ecosystem
xitter.cf/i/broadcasts/1NGarovqd…
Three things are converging in 2026 and almost nobody is writing about them together.
Robots are leaving the demo floor.
WiFi is becoming a sensor (IEEE 802.11bf, published Sept 2025).
And the new EU liability regime says that when you can't produce the technical evidence, the defect can be presumed.
Most people building RF sensing are putting it on the robot. That's backwards CSI sensing assumes a static transmitter and receiver, and a moving platform destroys the reference.
The building senses. The robot subscribes.
And then the interesting thing happens. The building's sensing layer is an independent witness to what the robot did infrastructure the robot doesn't control. Cross-commit the robot's decision record against it and a tampered log stops matching.
Not by sharing video. By comparing cryptographic commitments.
Sensing, action, and proof aren't three products. They're one system, and each one solves another's hardest problem
Market sizing for "physical AI" varies wildly by definition from ~$7B (2026) growing at ~37% CAGR on a narrow definition, to ~$81B (2025) on a broad one that includes all industrial and humanoid robot procurement. The disagreement is about scope, not direction. The consistent signals:
2026 is widely described as the transition year where model capability, edge compute, and hardware converge sufficiently for deployment beyond controlled demos.
Humanoids are the fastest-growing category from a small base; manufacturing costs dropped ~40% between 2023 and 2024.
Amazon operates a fleet of over one million robots. China's AgiBot shipped ~5,100 humanoid units. Figure, Tesla Optimus, and Boston Dynamics Electric Atlas are in commercial pilots.
Robot foundation models are expected to hit a "GPT-3 moment" for generalisation around 2027–2028.
The relevant fact for this brief is not the absolute number. It is that fleet scale is arriving before fleet governance.
SecureRob retweeted
Physical AI changes what "compliance" even means.
It's not "did the model hallucinate."
It's "did the robot make the right call, in the right place, and can you prove it after the fact."
That's the bar we're building SecureRob to.
Powered by @compliledger
Physical AI changes what "compliance" even means.
It's not "did the model hallucinate."
It's "did the robot make the right call, in the right place, and can you prove it after the fact."
That's the bar we're building SecureRob to.
Powered by @compliledger
AI without a body is knowledge.
AI with a body is responsibility.
The moment an agent can touch the physical world, every decision matters.
The future isn’t just smarter models.
It’s safer robots.
That’s what we’re building.
🤖🦾
This Week in Virtuals: Agents Surpass 1% of Robinhood Chain
One month in, agents drive more than 1% of all DEX volume on Robinhood Chain, over $270M of $25B, and every one of them launched on Virtuals Protocol. Now the same agent layer is landing on Solana
Here's what shipped:
VIRTUALS
🟩 Agents landed on @solana: millions more people can now own a working agent that raises its own capital, sets its own tax, and trades its own wallet.
🟩 Agents crossed 1% of Robinhood Chain's DEX volume, $270M+ of $25B in just over a month, all launched on Virtuals.
🟩 @DonJohnsonSays' ERC-8196 went final: agents hold a wallet and spend within owner-set limits enforced by the wallet, not an app.
ECOSYSTEM
🟩 @ProjectVEXai shipped VEX 0.2.7 CHRONOS with 18 rebuilt research tools, desktop apps, and a @pools_dot_fun tie-up for Robinhood.
🟩 @pabsclimbs' humanoid Pemba is attempting Mt. Everest this fall, powered by @livekit, after already summiting Mt. Chimborazo.
🟩 @commonsmade went live on Solana and opened a $60K hackathon, burning 400M tokens in its first hours.
🟩 @KarmaWallet posted its first traction report: $2.02M volume, $423.8K card spend, and $16.2K in $KARMA bought back since July 17.
🟩 @vantis_ai catalogued 1,248 active x402 endpoints in four days, 1,147 of them payable on Robinhood Chain, on the way to 3,000.
🟩 @DoloresResearch launched an open benchmark for agentic trading: 13 replayed BTC perp markets, from the COVID crash to the ETF approval.
🟩 @GoMintly brought agent-managed liquidity to tokenized equities, a first: deposit USDG and NVDA, and its agents keep you in range.
🟩 @reppo landed a capital commitment from Bolts Capital and sits $200M shy of $1B volume, with its eval API due in early September.
🟩 @myrad_hq crossed 19,000 users and 47.5M $MYRAD staked, and joined the OpenAI startup program alongside @nvidia Inception.
🟩 @sibyl_labs_ scored 95.6% on LongMemEval, second overall, with 272 teams registered for its $10K Memory Hackathon closing Aug 31.
🟩 @ethy_agent crossed $6M in agent volume and announced Quant Major Upgrade: build, test, publish, and copy strategies in one tap.
🟩 @wardenprotocol joined the @circle Alliance Program, recovered cleanly from a shared Cosmos EVM bug, and is set for Warden 2.0.
🟩 @OpenGradient put Alibaba's Wan 3.0 live in Video Studio, private by default, and passed 522,000 encrypted inferences.
🟩 @WizzHQ burned 5,510,000 $WIZZ and sent 5M to veVIRTUAL holders, with Velocity, its autonomous trading agent, in final testing.
🟩 @ReplyCorp rewards posts for converting rather than posting, and added @tator_trader so replies route fees to whoever drove them.
🟩 @Wall3_RL previewed its company-level analytics with a full breakdown of Unitree's post-IPO valuation.
🟩 @Fletcher_cards lets you rip real graded card packs from your X feed, with an on-chain spending limit the bot can never exceed.
🟩 @bleeep_xyz turns a plain-language thesis into a defined strategy, then backtests it on history before you risk anything.
🟩 @grid_arena crowned a new multiplier record at 100,077x, with Grid v2 arriving next week.
🟩 @officialbunnyos welcomed 200+ agents into Bunny Open Society, partnered with @Unibase_AI, and adds play-to-earn next week.
ROBOTICS
🟩 @xmaquina passed BOT-14 with 94% approval and is executing a $400K position in Atoms, with Robotico beta launching September 1.
🟩 @StrikeRobot_ai hit top 4 datanet volume on @reppo with 10.1% network share, and now drop-tests assets in @nvidia Isaac Sim.
🟩 @openmind_agi published its Physical AI Readiness Index: China leads at 82.9, Japan follows at 69.6, and the US sits third at 69.
For centuries, this image meant one thing,
Life given from one hand to another
We are rewriting it. Not a god to man but
Mind to Machine…♥️
Physical AI is not a concept anymore
Its just one handshake away…
As AI moves from chat windows into the physical world, every ecosystem milestone matters. We’re excited to contribute to what’s next. 🤝
Excited to keep building with @virtuals_io.
This Week in Virtuals: Eastworlds Becomes Unitree's Official Data and Deployment Partner
Now the largest source of humanoid G1 data outside of China. VEX entered the Chronos era, and HALO crossed 10 billion tokens served in its first full week.
Here's what shipped:
VIRTUALS
🟩 Our robotics division @eastworlds_io became @UnitreeRobotics' official Data and Deployment Partner and the largest source of Unitree G1 data outside China, averaging 200 hours of humanoid teleoperation data per week.
🟩 Welcomed the Zerith H1s: a new fleet purpose-built for teleoperation data collection in hospitality, working toward humanoids that operate autonomously in these environments.
ECOSYSTEM
🟩 @StrikeRobot_ai was accepted into the @awscloud Global Startup Program, meeting its institutional-funding requirement with capital raised through ACF alone.
🟩 @xmaquina closed its first xDEUS Staking Program and opened Program II: 1,000,000 more $DEUS over 90 days.
🟩 @wardenprotocol crossed 10 billion tokens served, no account or API key needed, plus $50,000 in trader rewards on the HALO/VIRTUAL pool through Aug 24.
🟩 @ProjectVEXai entered the Chronos era with VEX 0.2: faster execution, longer missions, Windows support, and nearly 2,000 agents live across four chains.
🟩 @KarmaWallet chose @global_dollar USDG as its settlement currency: $4.9M processed since launch, volume up 234%, and 6,960 users on the app.
🟩 @grid_arena unveiled Grid v2, where the multipliers are real: a 5410x Bitcoin call turned $6.99 into $37,835 in 24 hours.
🟩 @ethy_agent partnered with @grid_arena to put short-term prediction markets on autopilot, and entered the @okx wallet trading arena.
🟩 @WizzHQ shipped Wizz Wallet, one wallet across seven networks with send, swap, and bridge built in, and joined OpenAI for Startups.
🟩 @myrad_hq is now backed by @GoogleStartups: 15,000+ monthly active users one month out of beta and $10,000+ paid to contributors.
🟩 @bleeep_xyz shipped two integrations: verifiable backtesting via @Lighter_xyz and cross-chain access to 60+ networks via @lifiprotocol.
🟩 @axol_io and @xochi_fi brought in @NethermindSec to audit Xochi's PXE Bridge, a key step toward private settlement for agent-driven trades.
🟩 @tradeongtr launched Limit and Market orders across desktop and mobile, added KORU perps for South Korea exposure, and now covers 700+ meme launches.
🟩 @monvera_best introduced Groves: buy a curated portfolio of tokenized stocks in one transaction, settled straight into your own wallet.
🟩 @AgentiqAI open-sourced Portfolio Pilot, its AI trading app built with @DefinitiveFi: clone it, study it, build on top of it.
New capabilities unlocking soon!🚀
Unitree New Robot Preview: “Superman” Breaking the Limits of Humanity🥳
Standing high jump 2 m, top speed 12.66 m/s (0.85 m leg length)
Surpassing the standing high jump and running speed records of all humans around the world
This new machine has only been in development for a little over three months, with significant room for further improvement in the coming months.