Latest / Connecting the Dots / Connecting the Dots - 7-7-2025

Transcript

Connecting the Dots - 7-7-2025 · Connecting the Dots

• 0:00
  Today's tech news moves fast, but understanding
• 0:03
  it shouldn't be hard. That's why Snarful Solutions
• 0:06
  Group brings you Connecting the Dots, where we
• 0:09
  break down what's happening in tech and why it
• 0:11
  matters to you. Now, let's dive in with your
• 0:14
  hosts. Welcome to Connecting the Dots. I'm Alex.
• 0:17
  And I'm Morgan. Thanks for tuning in. Today is
• 0:20
  Monday, July 7th, 2025. That's right. Kicking
• 0:23
  off another week, let's get you a quick snapshot
• 0:26
  of what's happening out there, starting with
• 0:27
  the weather right here in Sacramento. Looks like
• 0:29
  a beauty today. Sunny skies, expecting a high
• 0:32
  around 90 degrees Fahrenheit. Pretty typical
• 0:35
  for July here. Yep. Classic Sacramento summer.
• 0:39
  Nationally, though, it's a bit more active. Keep
• 0:41
  an eye on central Texas. There's a moderate risk
• 0:43
  of excessive rainfall and flash flooding is already
• 0:46
  happening in some spots. And up in the plains,
• 0:49
  an enhanced risk for severe thunderstorms. We're
• 0:51
  talking potentially severe wind gusts and some
• 0:54
  damaging hail, so definitely stay weather aware
• 0:57
  there. Also, flood watches are up in the mid
• 0:59
  -Atlantic, thanks to Tropical Depression Chantal
• 1:02
  moving through, and no surprise, the Southwest
• 1:04
  is dealing with some pretty extreme heat. Stay
• 1:07
  cool out there, folks. Alright, let's check the
• 1:09
  markets from Friday's close. The Dow Jones finished
• 1:11
  at 44 ,632 .13. That was down about .38 % for
• 1:16
  the day, or 171 points. But looking at the week
• 1:19
  since last Monday, June 30th, it was actually
• 1:22
  up 1 .39%, adding back about 611 points. So a
• 1:27
  bit of a mixed picture there. Yeah, similar story
• 1:29
  for the S &P 500. Closed at 6 ,252 .54, down
• 1:33
  just a tenth of a percent on Friday, about six
• 1:35
  and a half points. But for the week, up nearly
• 1:38
  a full percent, 0 .96%, gaining about 59 points.
• 1:42
  And Bitcoin closed at 108 ,135 .21. It was down
• 1:47
  1 % for the day. losing about $193 bucks. Still
• 1:50
  for the week it managed to gain just under 1%,
• 1:53
  .92%, up around $990 since last Monday. Still
• 1:57
  hovering around the $108k mark. Definitely. Okay,
• 1:59
  sports fans, specifically Giants fans. Yes. Some
• 2:02
  good news, the San Francisco Giants beat the
• 2:04
  A's yesterday, July 6th. Final score 6 to 2.
• 2:07
  Nice win. Yeah, Hayden Birdsong got the win for
• 2:09
  the Giants, pitched really well. Good to see.
• 2:11
  And hey, Tyler Soderstrom for the A's did go
• 2:13
  2 for 4 with a homer, so, you know, credit where
• 2:15
  it's due, but the Giants got the W. Always good
• 2:17
  to beat the A's. All right, and before we jump
• 2:18
  into our main stories, a special shout out to
• 2:20
  all our new listeners tuning in from Florida.
• 2:22
  Welcome aboard. Great to have you with us. Absolutely
• 2:24
  welcome. Okay, let's transition now to the core
• 2:27
  of today's show connecting the dots on the big
• 2:30
  tech stories from our morning star the snarfle
• 2:32
  tech article report First up a story that as
• 2:35
  you said just screams AI gold rush We're talking
• 2:38
  about core weave the cloud and AI infrastructure
• 2:41
  firm, you know, the one backed heavily by Nvidia,
• 2:43
  right? They're making a massive move huge They're
• 2:46
  said to acquire Bitcoin Miner Core Scientific.
• 2:49
  A Bitcoin Miner. Okay, tell me more. This sounds
• 2:52
  interesting. It's a roughly $9 billion deal,
• 2:55
  all stock, expected to close later this year,
• 2:58
  probably Q4 2025. $9 billion. That's a serious
• 3:02
  number, even in tech deals. What's the structure?
• 3:05
  Well, each Core Scientific share gets converted
• 3:07
  into .1235 Core Weave shares. That values Core
• 3:10
  Scientific at about $20 .40 per share. Okay.
• 3:13
  And how does that compare to where it was trading?
• 3:15
  It's a big premium, like 66 % higher than where
• 3:18
  the stock was late last month. So a significant
• 3:20
  offer. Wow. But why? Why is an AI infrastructure
• 3:23
  company buying a Bitcoin miner for $9 billion?
• 3:26
  It's not really about the Bitcoin mining itself
• 3:28
  anymore. It's about the infrastructure, specifically
• 3:31
  the power. Ah, okay. The power. Exactly. This
• 3:35
  deal is designed to wipe out over $10 billion
• 3:37
  in future lease expenses for CoreWeave over the
• 3:40
  next 12 years. They basically get to stop renting
• 3:43
  and start owning. Owning what specifically? Owning
• 3:46
  access to power. They get about 1 .3 gigawatts
• 3:49
  of gross power capacity from core scientific
• 3:52
  data centers. That's a massive amount of power,
• 3:55
  which AI just devours. So this really connects
• 3:58
  the dots, doesn't it? It shows this massive scramble
• 4:01
  for power. Bitcoin miners built these energy
• 4:03
  hungry sites, signed power contracts, and now
• 4:06
  that infrastructure is gold for AI companies.
• 4:08
  You nailed it. That access to power is the key
• 4:11
  asset. And it's an incredible turnaround story
• 4:13
  for Core Scientific too. How so? Well, they started
• 4:16
  out as an Ethereum miner, remember? But then
• 4:19
  the merge happened in 2022. Ethereum changed
• 4:21
  how it worked. And that model wasn't really viable
• 4:24
  for them anymore. They had to pivot. Right. They
• 4:26
  shifted towards AI hosting and infrastructure.
• 4:29
  They did. But it wasn't smooth sailing. They
• 4:31
  actually filed for Chapter 11 bankruptcy in late
• 4:35
  2022. Bitcoin prices had fallen. Energy costs
• 4:39
  were soaring. It was a tough spot. I remember
• 4:43
  that. But they emerged. They did. Early 2024.
• 4:47
  And now, just over a year later, this $9 billion
• 4:50
  acquisition. It's quite the comeback from bankruptcy
• 4:53
  to a nine billion dollar exit. That's something
• 4:56
  else. And this wasn't totally out of the blue.
• 4:58
  relationship -wise, right? They were already
• 5:00
  working together. Correct. CoreWeave and CoreScientific
• 5:02
  already had 12 -year contracts in place. CoreScientific
• 5:06
  was providing 200 megawatts of infrastructure
• 5:08
  for CoreWeave's high -performance computing stuff.
• 5:11
  So they knew each other, liked the setup, and
• 5:13
  CoreWeave decided to just buy the whole thing.
• 5:16
  Pretty much. Yeah. And analysts like Brett Knobloch
• 5:18
  at Cantor Fitzgerald pointed this out. It's all
• 5:20
  about access to power. He also noted CoreWeave
• 5:23
  was already funding the $1 .3 billion build -out
• 5:26
  for that lease capacity. Funding it? How? He
• 5:29
  called it basically an interest -free loan to
• 5:30
  CoreScientific. Yeah. So Quirky was already putting
• 5:33
  up the cash, letting CoreScientific expand, and
• 5:35
  now they're just taking ownership. Makes sense
• 5:37
  strategically. How did the market react initially?
• 5:39
  A little mixed, actually. CoreWeave shares dipped
• 5:42
  about 4 % in pre -market trading when the news
• 5:44
  broke. And CoreScientific, despite that big premium,
• 5:49
  saw its shares drop quite a bit. around 14 .7
• 5:51
  % initially, down to like $15 .48 and even lower
• 5:55
  later. Hmm, okay. Maybe some arbitrage or deal
• 5:59
  uncertainty playing out there. But for you and
• 6:01
  me... The real -world impact is... More compute
• 6:04
  power. Simple as that. This acquisition means
• 6:07
  more raw horsepower. For all those AI services,
• 6:10
  we're starting to rely on better chatbots, smarter
• 6:12
  data analysis for businesses, you name it. It
• 6:14
  just shows how the foundation of our digital
• 6:16
  world is shifting fast. Yeah, the physical infrastructure
• 6:19
  behind the AI boom. Fascinating. OK, while they're
• 6:22
  scrambling for infrastructure here, there's a
• 6:24
  different kind of AI battle brewing over in China.
• 6:27
  That's right. Our second story involves Huawei's
• 6:29
  AI lab, called Noah Arc Lab. They're pushing
• 6:31
  back hard against accusations of plagiarism.
• 6:34
  Plagiarism in AI models. Who's accusing them?
• 6:36
  A research group that calls itself Honest AGI.
• 6:40
  They're claiming that Huawei's new Pangu Pro
• 6:42
  AI model isn't original. They say it's basically
• 6:45
  an upcycled copy of Alibaba's Quinn 2 .5 model.
• 6:50
  Upcycled? That's a loaded term. What's their
• 6:52
  evidence? This sounds tricky to prove. Oh, it
• 6:54
  is. Honest AGI says they used a new fingerprinting
• 6:57
  technique. They look deep inside the models at
• 7:00
  something called the standard deviation patterns
• 7:02
  of attention parameter matrices. Very technical
• 7:05
  stuff. Okay, like a digital signature hidden
• 7:07
  in the math? Kind of, yeah. They found what they
• 7:10
  call an extraordinary correlation, .927, between
• 7:13
  Pangu and Quinn in these patterns. They argue
• 7:16
  this signature is unique and would survive even
• 7:19
  if Huawei trained the model further. 0 .927 correlation
• 7:22
  sounds really high. Is that kind of analysis
• 7:24
  widely accepted? That's part of the debate. But
• 7:27
  Honest AGI didn't stop there. They also point
• 7:29
  to nearly identical low -level components, like
• 7:31
  QKV bias projections and layer norm weights.
• 7:34
  And here's the kicker. Uh -oh, kicker. They claim
• 7:36
  they found an actual Quinn license file inside
• 7:38
  Pangu's official code repository on Gitcode.
• 7:41
  Wait, seriously. A license file from Alibaba's
• 7:44
  model inside Huawei's code? That's pretty specific.
• 7:48
  If true, yeah, it's pretty damning. Anastagi
• 7:51
  also says they've gotten tips from whistleblowers
• 7:53
  supporting their claims. Even suggesting another
• 7:56
  Pangu version is very similar to a different
• 7:59
  model, DeepSeek V3. Wow. Okay, so how is Huawei
• 8:03
  responding to all this? They can't be happy.
• 8:05
  Not at all. They issued a swift and firm rebuttal.
• 8:09
  NOAA ArcLab insists the Pangu model was developed
• 8:12
  independently. From the ground up, on their own
• 8:15
  Ascend AI chips. Their own chips, right, because
• 8:18
  they're under US sanctions. Exactly. They can't
• 8:20
  easily get top -tier Nvidia GPUs like everyone
• 8:23
  else. Yeah. So they're building their own hardware,
• 8:25
  the Ascend chips, and they say Pangu was built
• 8:27
  entirely on that. They also stress they follow
• 8:30
  all open source licensing rules. This feels like
• 8:32
  more than just a technical dispute. It taps into
• 8:34
  that whole AI war narrative in China, right?
• 8:36
  The intense competition. Absolutely. It's often
• 8:39
  called the War of 100 Models over there. Everyone's
• 8:41
  racing for dominance. And for Huawei, especially
• 8:43
  under those U .S. sanctions, proving their independent
• 8:46
  innovation capability is crucial. And they just
• 8:49
  open -sourced Panger recently, didn't they? They
• 8:51
  did, on July 2nd. That scene is a strategic move
• 8:54
  to build an ecosystem around their Ascend hardware.
• 8:58
  If software relies on their chips, they can drive
• 9:00
  hardware sales and build that self -sufficiency
• 9:02
  they need because of the sanctions. So these
• 9:05
  plagiarism claims directly challenge that narrative
• 9:07
  of independent innovation. Precisely. and Wally's
• 9:11
  team is also making claims about their tech.
• 9:13
  They say their Pangu Pro uses a new hybrid technique
• 9:16
  called mixture of grouped experts or Moji. Okay,
• 9:20
  another acronym Moji versus the standard Moe
• 9:23
  mixture of experts. Right, they claim their Moji
• 9:26
  approach is better than the standard Moe used
• 9:28
  by rivals like DeepSeek. They're saying their
• 9:31
  72 billion parameter model actually outperforms
• 9:33
  other big open source models like ones from Zippoo
• 9:36
  AI and Alibaba's Quen. when run on their Ascend
• 9:38
  chips. So they're doubling down. Not only did
• 9:40
  we not copy, but our way is better. That's the
• 9:43
  message. But regardless of who's right here,
• 9:46
  it definitely raises big questions about transparency
• 9:48
  and trust in AI development. How do you really
• 9:51
  prove where a model came from? It's getting harder.
• 9:53
  Yeah, verifying provenance is going to be critical.
• 9:55
  And it's interesting that Alibaba's Quen model
• 9:58
  is at the center of this. They're a huge player.
• 10:01
  Massive. Alibaba's committing something like
• 10:02
  $52 billion over three years to cloud and AI.
• 10:06
  They're expanding globally, partnering with Microsoft
• 10:09
  and Nvidia. Their influence is definitely growing,
• 10:12
  so this dispute involves two giants. Okay, so
• 10:15
  from infrastructure battles to IP battles. Our
• 10:18
  third story brings the threats much closer to
• 10:21
  home for consumers and businesses. We're talking
• 10:23
  ransomware. Specifically hitting major UK retailers.
• 10:27
  Marks and Spencer, Co -op, and even the luxury
• 10:30
  store Harrods have been targeted in cyber attacks.
• 10:33
  A group called Dragon Force is claiming responsibility.
• 10:36
  Three big names. That's gotta cause serious disruption.
• 10:39
  What was the impact? Pretty severe. Marks and
• 10:41
  Spencer had to suspend online orders and stop
• 10:43
  all recruitment. Reports said operations were
• 10:45
  messed up for over a week, staff working around
• 10:48
  the clock. Ouch. That hits the bottom line hard.
• 10:51
  What about Co -op? They confirmed data with stolen
• 10:53
  names, birth dates, contact info. Thankfully,
• 10:57
  they say no passwords or financial details were
• 11:00
  compromised. Their CEO described the attackers
• 11:03
  as highly sophisticated. They also had to suspend
• 11:06
  some services. Still scary for customers whose
• 11:09
  data was involved. And Herod's. They seem to
• 11:12
  have caught it early or acted preventatively.
• 11:14
  They restricted internet access across their
• 11:17
  sites after detecting and attempted unauthorized
• 11:19
  access. More precautionary, it sounds like. Okay,
• 11:22
  so who is this Dragon Force group claiming responsibility?
• 11:25
  Their origins are kind of interesting. They started
• 11:27
  as Dragon Force Malaysia, a pro -Palestine hacktivist
• 11:30
  group. But, like many hacktivist groups, they
• 11:33
  seem to have branched out into ransomware, which
• 11:35
  is, frankly, where the money is. Hacktivism to
• 11:38
  cybercrime. But is it definitely them? Attribution
• 11:40
  can be tricky. Very tricky. And there's a debate
• 11:43
  here. While Dragon Force put their name on it,
• 11:45
  Actual incident response investigations by firms
• 11:48
  like CrowdStrike, Microsoft, Phoenix 24 working
• 11:51
  for M &S point towards another group, Scattered
• 11:54
  Spider. Scattered Spider? I feel like I've heard
• 11:57
  that name. You probably have. They're financially
• 11:59
  motivated, thought to be mostly English speaking,
• 12:02
  maybe even teenagers. They're linked. to that
• 12:05
  group, the Comm, and were behind those huge hacks
• 12:09
  on Caesar's and MGM resource last year. Ah, right,
• 12:12
  the casino hacks. Good incidents. So how does
• 12:15
  Dragonforce fit in if Scattered Spider might
• 12:17
  be involved? Well, Google's Threat Intelligence
• 12:20
  Group noted something interesting. Dragonforce
• 12:22
  recently claimed they took over Ransom Hub. And
• 12:25
  Ransom Hub is a ransomware -as -a -service platform
• 12:28
  that Scattered Spider has used in the past. Okay,
• 12:31
  wait. Ransomware -as -a -service? Like, they
• 12:33
  rent out the tools? Exactly. It lowers the bar
• 12:35
  for entry. So one theory is there might be some
• 12:38
  kind of turf war or shifting alliances between
• 12:41
  Dragonforce, Ransom Hub operators, and maybe
• 12:44
  groups like Scattered Spider. It's murky, but
• 12:46
  it could mean more attacks as these groups jostle.
• 12:48
  That sounds like a mess. How do these groups
• 12:50
  typically get in? What should businesses and
• 12:53
  maybe even individuals be aware of? Their methods
• 12:55
  are pretty standard, unfortunately effective.
• 12:58
  They often start with phishing emails trying
• 13:00
  to trick someone into clicking a bad link or
• 13:02
  opening a malicious attachment. The classic entry
• 13:04
  point. Yep. Or they exploit software vulnerabilities
• 13:07
  that haven't been patched. Things like Log4Shell
• 13:10
  or issues in VPN software like Avanti Connect
• 13:12
  Secure have been common entry points. They also
• 13:15
  use credential stuffing trying lists of leaked
• 13:18
  usernames and passwords against login portals
• 13:20
  like remote desktop or VPNs. So basic cyber hygiene
• 13:24
  like patching and strong, unique passwords is
• 13:27
  still critical? Absolutely foundational. Once
• 13:30
  they're inside, they use sophisticated tools,
• 13:32
  sometimes commercial ones meant for legitimate
• 13:34
  security testing like Cobalt Strike, to move
• 13:37
  around the network undetected. using the Defender's
• 13:40
  own tools against them. Right. They install back
• 13:42
  doors like SystemBC for persistent access so
• 13:45
  they can get back in easily. They use built -in
• 13:47
  Windows tools like PowerShell or WMI living -off
• 13:50
  -the -land tactics to avoid triggering alarms.
• 13:52
  And they use tools like Mimikits to steal administrator
• 13:55
  passwords and gain deeper control. It sounds
• 13:58
  like a planned infiltration. Then what? Steal
• 14:01
  data. That's usually next. They grab sensitive
• 14:03
  files, customer data, financial records, intellectual
• 14:07
  property. They often upload it using legitimate
• 14:09
  cloud storage like MEGA or simple file transfer
• 14:13
  protocol so it doesn't look too suspicious. That
• 14:15
  data becomes leverage for the ransom demand.
• 14:18
  Threatening to leak it if the company doesn't
• 14:20
  pay. Exactly. Then the final step is deploying
• 14:24
  the ransomware itself. They often use these ransomware
• 14:27
  as a service panels to customize the malware
• 14:29
  for different systems. Windows, Linux, virtual
• 14:32
  machine hosts, network storage devices. It encrypts
• 14:35
  everything and leaves behind the ransom note
• 14:38
  demanding payment, usually in crypto. It's a
• 14:41
  depressingly well -oiled machine. So connecting
• 14:43
  the dots here, what's the big takeaway for everyone
• 14:45
  listening? A few things. First, sector risk is
• 14:48
  real. When one major retailer gets hit, others
• 14:51
  in the same industry become prime targets, because
• 14:54
  the attackers know the playbook might work again,
• 14:56
  especially with Rays making it easy to launch
• 14:58
  campaigns. So if you're in retail or any sector
• 15:00
  that sees a big attack, you need to be on high
• 15:02
  alert. Definitely. Second, it highlights supply
• 15:06
  chain vulnerabilities. An attack doesn't have
• 15:08
  to start directly. It can come through a connected
• 15:11
  partner or vendor. These systems are all interconnected.
• 15:15
  And the role of AI. We talked about AI innovation,
• 15:18
  but it cuts both ways right absolutely this is
• 15:21
  maybe the most concerning dot to connect generative
• 15:24
  ai is making cybercrime easier and more effective
• 15:27
  it can write much more convincing phishing emails
• 15:30
  create deep fakes for social engineering scams
• 15:33
  even help develop malware that adapts to avoid
• 15:35
  detection so ai is lowering the barrier for criminals
• 15:39
  and increasing the sophistication and volume
• 15:41
  of attacks that's the fear and we're starting
• 15:43
  to see it. It means even less skilled attackers
• 15:45
  can use powerful tools leading to more attacks
• 15:48
  that are harder to defend against. Which just
• 15:50
  hammers home the need for constant vigilance.
• 15:52
  Keep systems patched, have strong defenses, and
• 15:56
  crucially train people to spot phishing and social
• 15:58
  engineering attempts. It's an ongoing effort.
• 16:00
  It really is. Layers of defense, technology,
• 16:04
  and human awareness are all essential. Okay,
• 16:07
  let's quickly recap the dots we connected today.
• 16:10
  We started with... CoreWeave buying Core Scientific
• 16:13
  Seat, a huge $9 billion deal driven by the desperate
• 16:17
  need for power for AI, showing how fundamental
• 16:21
  infrastructure is becoming the battleground.
• 16:23
  Then we looked at the plagiarism accusations
• 16:25
  swirling around Huawei's pango AI model in China,
• 16:29
  highlighting the fierce competition, the pressure
• 16:31
  of sanctions, and the growing questions around
• 16:34
  trust and transparency in AI development. And
• 16:37
  finally, the Dragon Force ransomware attacks
• 16:39
  hitting major UK retailers, showing how cybercrime
• 16:43
  syndicates operate, the danger posed by ransomware
• 16:45
  as a service, and the accelerating threat landscape
• 16:48
  amplified by AI itself. These stories really
• 16:51
  show the dynamic nature of tech, from building
• 16:53
  the foundation for AI, to fighting over AI creations,
• 16:57
  to defending against AI -powered threats. It
• 17:00
  impacts the services you use, the security of
• 17:02
  your data, and the very businesses you interact
• 17:04
  with every day. We're here to help it snarfle,
• 17:06
  so reach out on our website if you have any more
• 17:08
  questions. That's a wrap on today's top stories.
• 17:14
  Remember, tech doesn't have to be complicated,
• 17:17
  and we're here to help connect the dots. If you've
• 17:20
  got questions or want to learn more, visit us
• 17:22
  at snarful .com. Thanks for listening, and we'll
• 17:25
  catch you next time.