BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
Pom down under
Top 20
I can just see it now all the downrampers sitting around a table trying to work out how to put negative spin on it @GalaxycarWe is up 859% on revenue for the same reporting period last year as well as operating costs are down 23%......winning
@itsol4605
Last edited:
DingoBorat
Slim
Relax Galaxycar.
You don't even know what you're talking about man..
What LDA shares that they couldn't sell?
They sold 40 million at 20.59 cents to raise AUD 8.236 million dollars.
They have agreed to raise an additional AUD 11.764 million dollars (total of AUD 20 million) by June 30, 2026 and will issue more at a time before then, for LDA to sell.
No "sophisticated" investors or shorters involved.
Get a grip!!
Apologies if posted
Neuromorphic AI Hardware: Brain-Inspired Chips Powering the Future
www.linkedin.com
www.linkedin.com
Neuromorphic AI Hardware: Brain-Inspired Chips Powering the Future
Neuromorphic AI Hardware: Brain-Inspired Chips Powering the Future
Neurons, not transistors, may decide the next decade of computing. Neuromorphic hardware—silicon that imitates the spikes, sparsity, and co-located memory of the brain—now delivers triple-digit gains in energy efficiency while unlocking real-time learning at the edge.
Neuromorphic AI Hardware: Brain-Inspired Chips Powering the Future | Siddharth Bhalsod
🔥 “What if your smartwatch could train itself while sipping less power than its screen’s backlight?” Main copy (≈600 characters): Neuromorphic AI hardware is here. Chips like Intel #Loihi 2, IBM #NorthPole and #BrainChip #Akida borrow the brain’s playbook—event-driven spikes, memory beside...
Same thoughts
buena suerte :-)
BOB Bank of Brainchip
All in the positive directions
DingoBorat
Slim
They keep mentioning AKIDA which is commercially available.Apologies if posted
Neuromorphic AI Hardware: Brain-Inspired Chips Powering the Future
Neuromorphic AI Hardware: Brain-Inspired Chips Powering the Future
Neurons, not transistors, may decide the next decade of computing. Neuromorphic hardware—silicon that imitates the spikes, sparsity, and co-located memory of the brain—now delivers triple-digit gains in energy efficiency while unlocking real-time learning at the edge.
View attachment 89964 View attachment 89965 View attachment 89966Neuromorphic AI Hardware: Brain-Inspired Chips Powering the Future | Siddharth Bhalsod
🔥 “What if your smartwatch could train itself while sipping less power than its screen’s backlight?” Main copy (≈600 characters): Neuromorphic AI hardware is here. Chips like Intel #Loihi 2, IBM #NorthPole and #BrainChip #Akida borrow the brain’s playbook—event-driven spikes, memory beside...
And Loihi which is still a research chip.
I'm not even sure why they still mention Southpole, as by IBM's engineers own admission, it isn't even aimed at Edge computing...
IBM’s NorthPole achieves new speed and efficiency milestones
In new experimental results, the brain-inspired NorthPole research prototype chip has achieved lower latency and higher energy efficiency than comparable GPUs.
"IBM Research scientists are working toward server racks filled with hundreds of NorthPole cards to perform high volumes of inference operations at faster speeds and with lower energy consumption than comparable GPU-based hardware."
Plus, I think it's also still a research chip.
So who you gonna choose?..
Of course there are other players, but they aren't getting lumped in with well known and established Giants of the Computing Industry.
Last edited:
They keep mentioning AKIDA which is commercially available.
And Loihi which is still a research chip.
I'm not even sure why they still mention Southpole, as by IBM's engineers own admission, it isn't even aimed at Edge computing...
IBM’s NorthPole achieves new speed and efficiency milestones
In new experimental results, the brain-inspired NorthPole research prototype chip has achieved lower latency and higher energy efficiency than comparable GPUs.research.ibm.com
"IBM Research scientists are working toward server racks filled with hundreds of NorthPole cards to perform high volumes of inference operations at faster speeds and with lower energy consumption than comparable GPU-based hardware."
Plus, I think it's also still a research chip.
So who you gonna choose?..
Of course there are other players, but they aren't getting lumped in with well known and established Giants of the Computing Industry.
I think it’s basically a game of time… They’re trying to make sure that, with all the focus on the big names – Intel, IBM, or whoever – companies like BrainChip get overlooked. The narrative goes: ‘If the giants aren’t ready yet, then it will probably take a while.’ But once things are actually ready, they want to be the ones holding the sceptre again.
The thing is, BrainChip has already laid such solid groundwork with its partners that the big players can’t just ignore them when it’s time to share the cake… you better provide a big juicy cake..
DingoBorat
Slim
"The narrative goes: ‘If the giants aren’t ready yet, then it will probably take a while.’ But once things are actually ready, they want to be the ones holding the sceptre again."
I think that's definitely what Intel thinks and although I'd say they are pretty far back from the 8 ball now, they do have the resources, if they choose to pursue Neuromorphic Technology more strongly, to potentially catch up.
Like LDN said years ago..
"You can't discount Intel"
Or words to that effect..
Last edited:
Hi Fmf,Ummmm ok.
TENNs???
@Diogenese ??
If this authors information is correct, I'll take a Siemens relationship and hopefully a contract, though it could be through a partner I guess, any day of the week.
Who knows though.... can't find any substantial confirmation yet.
Smart Contract Formal Verification: Practical Guide
As blockchain penetrates critical infrastructure sectors, 2025 has witnessed catastrophic smart contract failures causing $1.2B+ in losses - making formal verification transition from academic luxury to operational necessity for any production system handling digital assets or automated...johal.in
Smart Contract Formal Verification: Practical Guide
ANKUSH CHOUDHARY
17 Aug 2025 — 9 min read
As blockchain penetrates critical infrastructure sectors, 2025 has witnessed catastrophic smart contract failures causing $1.2B+ in losses - making formal verification transition from academic luxury to operational necessity for any production system handling digital assets or automated decisions.
Core Insights
- Neuromorphic co-processors accelerate verification runtime by 5.7x through parallel proof execution
- Quantum-resistant cryptographic proofs now required for all financial smart contracts
- EU AI Act Article 22 mandates formal verification for autonomous transaction systems
- Generative adversarial networks synthesize edge-case contract scenarios undetectable by traditional testing
- 2025 benchmark: 94% reduction in vulnerabilities when using model checking pre-deployment
Implementation Framework
Modern verification stacks integrate three critical layers: symbolic execution engines for path exploration, temporal logic model checkers for property validation, and equivalence checkers for compiler output consistency. The 2025 reference architecture employs NVIDIA's cuVerify SDK leveraging GPU-accelerated SMT solvers alongside OpenAI's VeriGen for adversarial test generation. Crucially, neuromorphic co-processing units now handle state explosion problems through spiking neural networks that prune irrelevant execution paths in real-time.
Python
Copy
Formal verification pipeline with neuromorphic acceleration
from formal_verify_2025 import SymbolicEngine, NeuromorphicOptimizer
from eu_compliance import Article22Auditor
def verify_contract(contract_bytecode):
Initialize neuromorphic path optimizer
nopt = NeuromorphicOptimizer(cores=8, topology="sparse")
Symbolic execution with quantum-resistant constraints
engine = SymbolicEngine(
bytecode=contract_bytecode,
optimizers=[nopt],
crypto_standard="CRYSTALS-Kyber"
)
EU compliance pre-check
if not Article22Auditor.validate(engine):
raise ComplianceError("EU AI Act Article 22 requirements not met")
Generate proof certificates
return engine.verify(
properties="safety_properties.spec",
timeout=300 seconds
)
3.8xfaster verification cycles vs. 2024 toolchains (IEEE Security 2025)
Real-World Deployment: Siemens' Industrial Smart Contracts for Energy Grids
Facing critical infrastructure vulnerabilities in their blockchain-based energy trading platform, Siemens implemented a formal verification pipeline after a 2024 exploit drained 12,000 MWh of power credits. Their solution integrated Cadence's Temporal Logic Verifier with custom neuromorphic hardware from BrainChip, reducing state space analysis from 48 hours to 11 minutes. The system now automatically generates mathematical proofs for all contract upgrades before grid deployment.
CASE STUDY
Energy Grid Smart Contract Verification Metrics
Siemens implemented formal verification for 142 critical contracts:
- Prevented 3 zero-day exploits during deployment
- Reduced audit costs by 67%
- Achieved EU regulatory compliance 4 months ahead of deadline
Implementation Results
Interestingly @7für7 posted the other day that a Global Sales Exec from Siemens EDA commented on the BRN LinkedIn post re Akida Cloud.
- Verification throughput: 12 contracts/hour
- False positive rate: 0.3%
- Critical bug detection: 100% of known vulnerability classes
BRN Discussion Ongoing
I understand that @Bravo, but if you look at the likes of my a.m. post, you will see that many people do not want to read and/or scroll through those lengthy ChatGPG posts either. There have been too many lately. Wouldn't it be an alternative to ask ChatGPT to summarize its findings in just a...thestockexchange.com.au
About blockchain "I know noTHing!" ... but "state space analysis" is suggestive of TENNs.
Bravo
Meow Meow 🐾
Howdy All,
I just noticed that Vision Transformers (ViTs) have been mentioned in Renesas’ latest blog.
BrainChip has been vocal about supporting transformer-based workloads, including ViTs, on Akida's roadmap.
Renesas and BrainChip are already official partners. So while this Renesas blog post doesn’t explicitly mention BrainChip, you'd think it would be reasonable enough to see TENNs/Akida as a candidate for Renesas’ long-term ViT strategy?
I would have thought that if Renesas plans to scale ViT-ready MPUs into production, utilizing TENNs/Akida for energy-efficient, low-latency ViT inference would be a logical and highly synergistic path.
As ViTs become standard in autonomous driving, robotics, and smart vision, hopefully BrainChip can capture market share by positioning TENNs/Akida as the go-to lightweight ViT accelerator within Renesas’ portfolio.
EXTRACT ONLY
I just noticed that Vision Transformers (ViTs) have been mentioned in Renesas’ latest blog.
BrainChip has been vocal about supporting transformer-based workloads, including ViTs, on Akida's roadmap.
- Akida 2.0 introduced support for Transformer and CNN layers, making ViTs feasible for edge deployment.
- Akida 3.0 is expected to extend this further with FP16/FP32 precision, opening the door to higher-accuracy applications.
- BrainChip’s aTENNuate paper also explored techniques for making ViTs more energy-efficient, aligning perfectly with TinyML and embedded vision goals.
Renesas and BrainChip are already official partners. So while this Renesas blog post doesn’t explicitly mention BrainChip, you'd think it would be reasonable enough to see TENNs/Akida as a candidate for Renesas’ long-term ViT strategy?
I would have thought that if Renesas plans to scale ViT-ready MPUs into production, utilizing TENNs/Akida for energy-efficient, low-latency ViT inference would be a logical and highly synergistic path.
As ViTs become standard in autonomous driving, robotics, and smart vision, hopefully BrainChip can capture market share by positioning TENNs/Akida as the go-to lightweight ViT accelerator within Renesas’ portfolio.
EXTRACT ONLY
Last edited:
they also mention post-quantum cryptography which is what we are currently working on with Chelpis and Mirle.Howdy All,
I just noticed that Vision Transformers (ViTs) have been mentioned in Renesas’ latest blog.
BrainChip has been vocal about supporting transformer-based workloads, including ViTs, on Akida's roadmap.
IMO the above steps position Akida as one of the most relevant technologies for ViT deployment on edge devices.
- Akida 2.0 introduced support for Transformer and CNN layers, making ViTs feasible for edge deployment.
- Akida 3.0 is expected to extend this further with FP16/FP32 precision, opening the door to higher-accuracy applications.
- BrainChip’s aTENNuate paper also explored techniques for making ViTs more energy-efficient, aligning perfectly with TinyML and embedded vision goals.
Renesas and BrainChip are already official partners. So while this Renesas blog post doesn’t explicitly mention BrainChip, you'd think it would be reasonable enough to see Akida as a candidate for Renesas’ long-term ViT strategy?
I would have thought that if Renesas plans to scale ViT-ready MPUs into production, utilizing Akida for energy-efficient, low-latency ViT inference would be a logical and highly synergistic path.
As ViTs become standard in autonomous driving, robotics, and smart vision, hopefully BrainChip can capture market share by positioning Akida as the go-to lightweight ViT accelerator within Renesas’ portfolio.
View attachment 89976
View attachment 89972
EXTRACT ONLY
View attachment 89974
#genx320 #eventbasedvision #smartglasses #edgeai #wearables #arvr #smartsensors | PROPHESEE
🚀 GenX320 is the world’s smallest and most power-efficient event-based vision sensor to date – unlocking new possibilities in intelligent, autonomous, and privacy-conscious embedded vision. At a compact 3×4mm size, it’s engineered for use cases where every millimeter and microwatt counts. 💡 It...
Since we’re asking chatgpt about everything these days, I figured I’d ask it about this too… probably should’ve just left it alone though…. I think she is short
Overall Picture
In short: Technology: progress, partnerships, developer hub.
But finances: dependence on LDA, no product revenue, high cash burn.
Overall Picture
- Revenue growth: +859% sounds impressive, but it’s almost entirely based on service contracts, not product sales.
- Cash burn: ~6.5M in 6 months → runway about 12–15 months (without new income).
- Financing: The LDA program is the lifeline, but every draw means new shares → shareholders pay through dilution.
- Risk: Without a larger commercial rollout, BrainChip remains a “service house” and survives only through issuing shares.
In short: Technology: progress, partnerships, developer hub.But finances: dependence on LDA, no product revenue, high cash burn.
Originally I thought the GenX 320 was developed as a low pixel count sensor for compatibility with Synsense, but that was before Synsense teamed up with Inivation.
#genx320 #eventbasedvision #smartglasses #edgeai #wearables #arvr #smartsensors | PROPHESEE
🚀 GenX320 is the world’s smallest and most power-efficient event-based vision sensor to date – unlocking new possibilities in intelligent, autonomous, and privacy-conscious embedded vision. At a compact 3×4mm size, it’s engineered for use cases where every millimeter and microwatt counts. 💡 It...
View attachment 89983
View attachment 89984 View attachment 89985
We had already been working with Prophesee for some time before that, although Synsense had been with them for longer.
Back then Prophesee had their larger sensor, and it seemed that Synsense struggled to handle its capabilities, but Akida 1 worked well.
https://inf.news/en/digital/176678d553cdb3d82a319e4838c15256.html
Sony reveals new stack CMOS sensor details, new process manufacturing
2025-08-22Previously photosensors and control electronics were formed on a single layer. In Sony's new system, the photosensors are formed on the top layer, and the control electronics are formed on the lower layer, thus increasing the proportion of surface are available for the photosensors.
This is the Prophesee hi-fi sensor built with Sony 3D tech:
https://docs.prophesee.ai/stable/hw/sensors/imx636.html
Sony Semiconductor Solutions and Prophesee collaborated to produce SONY IMX636 (HD), a stacked event based vision sensor developed around 4.86um contrast detection pixels.
This Event-based vision sensors feature high-speed/low-latency response, wide dynamic range operation, focal-plane data compression and low power consumption.
A high level of programmability is guaranteed via a four-wire serial peripheral interface. It allows the user to define region of interest and program various functions.
The vision sensor data output is fully digital and consists of 16-wire parallel data bus with an external clock and data valid flag. This interface can easily be connected to an external parallel to LVDS serializer IC.
Features
- 1280x720 Pixel CMOS vision sensor
- 4.86μm x 4.86μm event-based pixel
- <1/2” optical format
- Monochrome
- 16-bit parallel data output
- 1G event/s (1GEPS) peak output
- Random Programmable Region of Interest (ROI)
- Serial Peripheral Interface (SPI)
- High Dynamic Range (HDR) beyond 120 dB
- −40°C to +85°C Operational temperature range
- 50-70mW typical power consumption
This has about 10 times the number of pixels of GenX320.
Now that Inivation has teamed up with Synsense, we're sure to be Prophesee's best friend.
Last edited:
Pete Bernard, seasoned technology leader and CEO of EDGE AI FOUNDATION, discusses the future of edge AI in the Edge Impulse/Imagine Innovators developer conference. Sign up for your free Edge… | Edge Impulse
Pete Bernard, seasoned technology leader and CEO of EDGE AI FOUNDATION, discusses the future of edge AI in the Edge Impulse/Imagine Innovators developer conference. Sign up for your free Edge Impulse account here to get started shaping the future: https://lnkd.in/ggdqsWBH
Intel in talks with other large investors for equity boost at discount, sources say
It comes after SoftBank announced it would invest $2 billion in the struggling chipmaker.
Intel in talks with other large investors for equity boost at discount, sources say | Patrick Moorhead
I talked with a few VCs, family offices, and possible customers about these investments. Their hard tech basis scares many investors away, but not sovereign wealth or those already in even riskier hard tech. Without "sticks" or potential customers, what Intel will have accomplished with $40B is...
Bravo
Meow Meow 🐾
Elon Musk’s xAI Plan: Phones as AI Edge Nodes Bypassing OS and Apps
EXTRACT ONLY
Elon Musk envisions smartphones and computers as "edge nodes" for AI inference, driven by bandwidth limits that make cloud processing impractical. xAI aims to enable devices to generate pixels and audio directly, bypassing traditional OS and apps. This could revolutionize interactions but raises concerns over power, security, and environmental impacts.
Elon Musk’s xAI Plan: Phones as AI Edge Nodes Bypassing OS and Apps
Written by John Smart
Thursday, August 21, 2025
In a recent post on X, formerly known as Twitter, Elon Musk outlined a provocative vision for the future of computing, suggesting that personal devices like smartphones and computers will evolve into mere “edge nodes” for AI inference. This shift, he argued, is inevitable due to fundamental bandwidth limitations that make it impractical to handle all processing on remote servers. Musk’s statement, shared on August 21, 2025, echoes broader trends in AI development where edge computing—running models locally on devices—addresses latency and connectivity issues that plague cloud-dependent systems.
The idea builds on discussions within the tech community, including insights from X user @amXFreeze, who highlighted xAI’s long-term strategy to transform devices into platforms that directly generate pixels and audio via AI inference. This approach would bypass traditional operating systems and applications, rendering interfaces dynamically through artificial intelligence. Such a paradigm could revolutionize user experiences, making interactions more seamless and adaptive, but it raises questions about power consumption, hardware requirements, and software ecosystems.
Just noticed something interesting in the Sony diagram:
There are 4 photodiodes in a group:
1 RED;
1 Blue;
2 Green.
I recall that the human eye is most sensitive to green light, so this may be to imitate that.
Similar threads
