Bryce Kunz
banner
tweekfawkes.bsky.social
Bryce Kunz
@tweekfawkes.bsky.social
51 followers 3 following 1.5K posts
AI, Cloud, & Cybersecurity focused posts. SaintCon, ApresCyber, BSides SLC, Cache, & RedRocks committee member. Helping the community grow via giving back!
Posts Media Videos Starter Packs
Windows 11's Voice Access tool for mobility disabilities now understands natural language—a very cool AI application. Instead of a rigid "Open Edge" command, you can now say "Can you open the Edge browser?" and it just works. #AI

archive.is/Ho86t
Just when you thought corporate job titles couldn't get any stranger, companies are now hiring for roles like "Vibe Growth Manager" and have experimented with a "Chief Vibe Officer" to encourage AI adoption. What a time to be alive. #Tech

archive.is/rSx8Q
Nvidia's new "Spark" machine is being called the world's smallest AI supercomputer. It’s tiny, but powerful enough to handle AI models with up to 200 billion parameters. The future of AI development might be smaller than we think. #AI

archive.is/5Vm8U
Turns out, larger AI models like GPT-4 benefit *more* from simple creativity-boosting prompts than smaller ones. This suggests a huge well of latent creativity in top-tier AIs we're just learning to tap. Fascinating stuff. #AI

archive.is/hEBJL
Mind-bending thought: What happens when you plug a large language model's reasoning ability directly into a constantly updated model of the physical world? We're finding out now, as Google connects Gemini to data from over 250 million places on Maps. #AI #Cybersecurity

archive.is/ecCM3
Peek behind the curtain of an AI job interview: it's not about the perfect answer, but *how* you think. A Stanford prof confirms they value a candidate's reasoning and ability to question assumptions above all. #Tech #Interviews

archive.is/qklWX
The strange paradox of the AI era: many of the same people who are skeptical of AI's impact are also regular users. The article notes that while students worry, over a third of them still use ChatGPT. We contain multitudes.

#AI #Tech
archive.is/3OHPu
AI is moving out of giant data centers and into our pockets and wearables. To get there, it needs a new kind of brain. Google's Coral NPU is an open blueprint for a tiny, powerful, and private AI processor meant for the devices we carry every day. #AI

archive.is/Qdpw6
A new AI model is >2x faster & 1/3 the price of a top model from just 5 months ago, with the same quality. Imagine if your phone's performance doubled every few months for a fraction of the cost. That's the pace of AI development now. #AI archive.is/awxlO
Ever wonder what a cutting-edge AI's "brain extension" looks like? It's not some arcane code. For Anthropic's Claude, a new capability can start with something as simple as a text file named "SKILL.md" in a folder. A powerful concept built on a simple foundation. #AI

archive.is/i1r9R
Think geopolitics is just for the news? Think again. A new PwC survey found that 60% of business leaders are increasing their cybersecurity investment as a direct response to global political uncertainty. The world stage is officially a cyber stage. #Cybersecurity

archive.is/VyuOU
What's the real weak link in security tech? Turns out it's not the tech at all. It's the gap between having an incident response plan and having a team that has actually practiced it until it becomes a reflex. #AI #Cybersecurity

archive.is/vbetd
Mindset shift: Good cybersecurity isn't about expertise, but accepting small inconveniences for big security gains. Think of updates & MFA like a digital flu shot—a small step that prevents a massive future problem. #Cybersecurity

archive.is/2pU1M
It's getting harder to land a junior role in cybersecurity. It's not just AI—companies are more demanding & the bar is higher. The biggest challenge for new grads isn't just learning the tech, it's getting a foot in the door. #CyberCareers #Tech

archive.is/nSBom
The article coins the term "Dr. Deepfake" to symbolize how cybercriminals are using AI to impersonate voices, faces, and full identities. It's a clever way to think about this new, faceless threat that weaponizes our senses. #Deepfake #TechNews

archive.is/xR0Dj
The issue with critical infrastructure isn't just that it can be hacked. It's that operators often don't know the root cause of an incident. This "cybersecurity blind spot" makes it incredibly hard to fix vulnerabilities and prevent future attacks. #AI

archive.is/9z4i0
As AI starts generating games and virtual worlds, some developers have a great reminder of the human element: they say the real challenge isn't the math or the code, but the "creative vision" behind it all.

archive.is/hbqsd
It's kind of wild that the tools we use for work are developing personalities. Slack says they're rebuilding Slackbot "from the ground up as a personalized AI companion," not just a tool. A subtle but huge shift in how we'll interact with software.

archive.is/XCGZg
Bigger than just tech: An analyst warns that massive spending on short-lived AI data centers could escalate an "industry crisis" into a "national economic crisis." Shows how deeply interconnected it all is.

archive.is/xtboN
It seems the biggest AI fans are the ones who understand it the least. This research points to a fascinating trend: expertise in AI often leads to increased skepticism, not more hype. Makes you think. #AI

archive.is/Sndrc
Fascinating AI fact: Whether an LLM has 600M or 13 BILLION parameters, it can be backdoored with just ~250 poisoned documents. A model's massive size and vast training data offer no extra protection from this specific attack.

archive.is/tLRUW
AI isn't just a tool; it's an amplifier. For people who already sent rambling emails or did half-hearted work, generative AI just lets them do more of it, faster. This "workslop" then pollutes the whole team's knowledge pool. Huh. #Cybersecurity #AI

archive.is/CjrQp
The emerging AI "double bind" for the average worker: 1) They're less equipped to leverage it strategically, but 2) even when they *do* produce outstanding AI-assisted work, they're less likely to get the career-advancing credit for it. #AI #Career

archive.is/vT2yw
The pace of AI progress is staggering:

Research shows AI's ability to handle complex tasks doubles every 7 months. At that rate, an AI could potentially complete an intern's 40-hour work week of tasks by late 2027.

#AI #Tech
archive.is/h1gF5
The next big AI breakthrough might not come from an algorithm alone. This partnership between OpenAI and Broadcom shows the frontier is moving towards co-designing the AI models and the physical chips together, from the ground up. The hardware and software are merging. #AI

archive.is/Wd9fH