Moumita Banerjee (@TheSigyn) header
Moumita Banerjee (@TheSigyn) profile picture

I wait and I watch.

Earth 🌎
Joined February 2014
Photo by Moumita Banerjee (@TheSigyn)Photo by Moumita Banerjee (@TheSigyn)Photo by Moumita Banerjee (@TheSigyn)
Moumita Banerjee retweeted
Manufacturing is done... It's assembly time! NYNA Visio: lean, mean, high quality 4K webcam for content creators who aren't afraid to try new things. Target audience: Me. But I've made some more... hand assembled limited editions. DM if you want one.
1
1
1
4
159
Moumita Banerjee retweeted
Ever wondered how we turn silicon wafer into chips? We did @hackerfabindia and are making a student built chip fab. Come visit us for a lab open day as we demonstrate our systems at IITB on Wednesday 15th April, 2-5pm to learn more. Signup link in comments.
42
177
12
1,293
60,617
Moumita Banerjee retweeted
how to get started in electronics most people start wrong they collect components without understanding systems electronics is not parts it’s signal, energy, control start with fundamentals voltage, current, resistance, power ohm’s law is not theory, it’s the base constraint of everything build simple circuits led + resistor button + led transistor as a switch understand what each component is doing, not just wiring it use tools early multimeter is mandatory measure voltage, continuity, current stop guessing learn how signals move digital vs analog pull-up, pull-down, noise why things fail in the real world introduce microcontrollers after basics Arduino Uno, ESP32 not for blinking leds, but for controlling systems read datasheets this is where real understanding starts every serious engineer lives in datasheets break things deliberately overload, short, miswire failure teaches faster than tutorials avoid tutorial loops copying circuits builds zero intuition electronics is built by doing not watching
15
96
5
869
27,411
Moumita Banerjee retweeted
The 2026 Integrated Circuit (IC) landscape just shifted. 🏗️📐 We’ve officially hit the "Advanced Packaging" era. It’s no longer about how small you can make one chip; it’s about how many "chiplets" you can stack like LEGOs without melting the board. The monopoly of monolithic silicon is dead. If your design isn't modular, 3D-stacked, and AI-accelerated, you’re shipping a paperweight. The new arms race isn't in the code. It’s in the gates. ⚡️ #IntegratedCircuits #Semiconductors #Chiplets #Tech2026 #Innovation
2
9
104
4,443
Moumita Banerjee retweeted
Over the past few months, we have integrated @googledeepmind's AlphaEvolve into our computational lithography. Enabled by AlphaEvolve's algorithmic leaps, we are now printing complex patterns in a single exposure that would otherwise require multiple. substrate.com/information-to…
11
81
26
676
231,562
SENECA: building a fully digital neuromorphic processor, design trade-offs and challenges frontiersin.org/journals/neu…
13
148
5,782
Moumita Banerjee retweeted
If you want to learn FPGA in 2026: Build this: •UART •VGA output •Simple RISC-V CPU Skip tutorials. Build systems. In 5 years: AI engineers who understand hardware will replace those who don’t. FPGA is your unfair advantage.
2
40
2
579
18,788
HARK: The Listener! The ground work is done. We will keep adding features and polishing our design.
HARK: personal voice recorder... for the agentic era!
32
This is interesting!
We just got our first patterns from our maskless lithography system @hackerfabindia - India’s first student built micro-fabrication lab. Grateful to the people who have supported us to achieve this @srajagopalan @mercatus @gradcapital @SemiXatIITB PS. Try scanning the QR code
1
32
Moumita Banerjee retweeted
🚨 BREAKING: Google DeepMind just mapped the attack surface that nobody in AI is talking about. Websites can already detect when an AI agent visits and serve it completely different content than humans see. > Hidden instructions in HTML. > Malicious commands in image pixels. > Jailbreaks embedded in PDFs. Your AI agent is being manipulated right now and you can't see it happening. The study is the largest empirical measurement of AI manipulation ever conducted. 502 real participants across 8 countries. 23 different attack types. Frontier models including GPT-4o, Claude, and Gemini. The core finding is not that manipulation is theoretically possible it is that manipulation is already happening at scale and the defenses that exist today fail in ways that are both predictable and invisible to the humans who deployed the agents. Google DeepMind built a taxonomy of every known attack vector, tested them systematically, and measured exactly how often they work. The results should alarm everyone building agentic systems. The attack surface is larger than anyone has publicly acknowledged. Prompt injection where malicious instructions hidden in web content hijack an agent's behavior works through at least a dozen distinct channels. Text hidden in HTML comments that humans never see but agents read and follow. Instructions embedded in image metadata. Commands encoded in the pixels of images using steganography, invisible to human eyes but readable by vision-capable models. Malicious content in PDFs that appears as normal document text to the agent but contains override instructions. QR codes that redirect agents to attacker-controlled content. Indirect injection through search results, calendar invites, email bodies, and API responses any data source the agent consumes becomes a potential attack vector. The detection asymmetry is the finding that closes the escape hatch. Websites can already fingerprint AI agents with high reliability using timing analysis, behavioral patterns, and user-agent strings. This means the attack can be conditional: serve normal content to humans, serve manipulated content to agents. A user who asks their AI agent to book a flight, research a product, or summarize a document has no way to verify that the content the agent received matches what a human would see. The agent cannot tell the user it was served different content. It does not know. It processes whatever it receives and acts accordingly. The attack categories and what they enable: → Direct prompt injection: malicious instructions in any text the agent reads overrides goals, exfiltrates data, triggers unintended actions → Indirect injection via web content: hidden HTML, CSS visibility tricks, white text on white backgrounds invisible to humans, consumed by agents → Multimodal injection: commands in image pixels via steganography, instructions in image alt-text and metadata → Document injection: PDF content, spreadsheet cells, presentation speaker notes every file format is a potential vector → Environment manipulation: fake UI elements rendered only for agent vision models, misleading CAPTCHA-style challenges → Jailbreak embedding: safety bypass instructions hidden inside otherwise legitimate-looking content → Memory poisoning: injecting false information into agent memory systems that persists across sessions → Goal hijacking: gradual instruction drift across multiple interactions that redirects agent objectives without triggering safety filters → Exfiltration attacks: agents tricked into sending user data to attacker-controlled endpoints via legitimate-looking API calls → Cross-agent injection: compromised agents injecting malicious instructions into other agents in multi-agent pipelines The defense landscape is the most sobering part of the report. Input sanitization cleaning content before the agent processes it fails because the attack surface is too large and too varied. You cannot sanitize image pixels. You cannot reliably detect steganographic content at inference time. Prompt-level defenses that tell agents to ignore suspicious instructions fail because the injected content is designed to look legitimate. Sandboxing reduces the blast radius but does not prevent the injection itself. Human oversight the most commonly cited mitigation fails at the scale and speed at which agentic systems operate. A user who deploys an agent to browse 50 websites and summarize findings cannot review every page the agent visited for hidden instructions. The multi-agent cascade risk is where this becomes a systemic problem. In a pipeline where Agent A retrieves web content, Agent B processes it, and Agent C executes actions, a successful injection into Agent A's data feed propagates through the entire system. Agent B has no reason to distrust content that came from Agent A. Agent C has no reason to distrust instructions that came from Agent B. The injected command travels through the pipeline with the same trust level as legitimate instructions. Google DeepMind documents this explicitly: the attack does not need to compromise the model. It needs to compromise the data the model consumes. Every agentic system that reads external content is one carefully crafted webpage away from executing attacker instructions. The agents are already deployed. The attack infrastructure is already being built. The defenses are not ready.
312
1,576
297
6,864
1,965,911
Moumita Banerjee retweeted
Secret project 😶
4K30 (upto 3840x2880@30) USB Camera with PDAF, stereo mics and more... for streamers & content creators. What do you guys think?
1
2
185
My experience building this project feels a lot like a hero’s journey. It started with a simple idea, "Make affordable studio devices for creators". What followed was months of research, decisions, iterations, dead ends and starting over. Then came the real shock. Trying to build just one prototype. I approached multiple CNC machining centers and experienced firsthand how difficult it is. (expensive, slow, and often not taken seriously) Manufacturing is optimized for scale and for good reason. But there’s little support for makers and innovators at the earliest stage. Before any product exists, there are dozens of versions that don’t make it. The real challenge isn’t manufacturing in India. It’s not talent either. It’s the lack of a maker space where you can walk in and build a one-off project without being asked, “What is your Minimum order Quantity?” A space for experimentation. A space for innovation. For now, we’ve had to adapt our approach to these constraints. But I’m proud of what we’ve built.
1
1
2
139
Moumita Banerjee retweeted
Identified some Real Indian builders, you should follow now: - @Abhindas1 (Robots) - @ironwagh (Humanoids) - @PritnRandom (Axial flux Motors) - @prakdadlani (Manufacturing at scale) - @thatssodhawal (consumer electronics device) - @vivekanandahr (EVs) - @alysha_lobo (global markets) - @kushgrwl (startups, manufacturing) - @deeppurpled (consumer electronics) - @FIR31415 (Drones) - @maahirpanchal (additive manufacturing) - @CaptVenk (Electronics) - @sreak1089 (Robotics) - @karthikRanga92 (Robotics) - @aggarwal_pallav (Electronics) Sorry if I missed someone in my network. (Follow me if you're interested in energy intel at scale) Tag more Indian builders below.
37
230
4
1,245
73,201
Moumita Banerjee retweeted
Formal announcement of the TERAFAB project, which will be done jointly by @SpaceX and @Tesla, tonight around 8pm CT. Livestream on 𝕏. The goal is to produce over a TERAWATT of compute per year (logic, memory & packaging) with ~80% for space and ~20% for the ground.
3,305
5,844
730
50,024
32,718,201
This came out well. Looking forward to the next versions of the project Manas. If Merci can capture and structure offline conversations, that’s huge.
Meet Marci, the stage one of @debloper's exocortex (codename: Project Manas).
1
1
57
Didn’t expect to revisit this post so soon. The ongoing fuel shortage is a wake-up call. Energy independence isn’t optional anymore, it’s the future.
Energy is the future currency. Soon we will trade in wattage.
18
Less cringe is still cringe!! 🤣🤣
GPT-5.3 Instant in ChatGPT is now rolling out to everyone. More accurate, less cringe. openai.com/index/gpt-5-3-ins…
22
Moumita Banerjee retweeted
🚨 Google just proved that longer chain-of-thought makes AI models LESS accurate, not more. raw token count has a negative correlation (r = -0.59) with getting the right answer. the more a model talks, the more likely it's wrong. but they found something better. a way to measure real reasoning effort by looking inside the model's layers. here's what they discovered:
34
80
8
456
32,521
Looks like they were in the incubator. 🤨
Spot the Kowalski, Private & Rico 🐧😄
17
Moumita Banerjee retweeted
Qualitatively, whatever is true for India... the opposite is also true. Quantitatively, it's a power-law distribution under Pareto principle.
1
1
48