Latest / Connecting the Dots / From Math-Solving Machines to Market Missteps: This Week in Tech and Trade

Transcript

From Math-Solving Machines to Market Missteps: This Week in Tech and Trade · Connecting the Dots

• 0:00
  Today's tech news moves fast, but understanding
• 0:03
  it shouldn't be hard. That's why Snarfl Solutions
• 0:06
  Group brings you Connecting the Docs, 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 Docs. I'm Alex.
• 0:17
  And I'm Morgan. It's Friday, November 7th, 2025,
• 0:21
  and we're coming to you from Sacramento, California.
• 0:23
  We are indeed shaking off that, uh... patchy
• 0:27
  morning fog here it's actually clearing up nicely
• 0:29
  for a sunny day high around 71 degrees yeah pretty
• 0:32
  nice here but uh if you're elsewhere especially
• 0:36
  near that strong storm system coming out of the
• 0:38
  pacific northwest uh definitely brace yourselves
• 0:41
  looks like wind and rain tracking across the
• 0:43
  country this weekend maybe a good weekend to
• 0:44
  stay inside perhaps listen to us talk tech and
• 0:46
  markets sounds like a plan Speaking of markets.
• 0:49
  Well, we saw a bit of turbulence this week, didn't
• 0:51
  we? We did the Dow Jones closed down about a
• 0:53
  hundred and eleven points yesterday and the whole
• 0:56
  week Saw a drop of over a thousand points So
• 1:00
  yeah some headwinds there. Mm -hmm. And similarly
• 1:02
  the S &P 500 it finished the week down what over
• 1:05
  3 % Bitcoin also took a pretty noticeable hit
• 1:08
  down about 6 ,500 bucks over the last seven days
• 1:11
  It's good reminder though that these daily and
• 1:13
  weekly shifts while we track them sometimes the
• 1:18
  bigger picture, the fundamental forces like the
• 1:20
  tech we're about to talk about are really driving
• 1:22
  things long -term. Exactly, those fundamentals
• 1:24
  are key and today we've got Three stories that
• 1:27
  really dig into where technology is feeling the
• 1:30
  pressure. Right. We're looking at the really
• 1:32
  abstract math behind it all. Then the tiny basic
• 1:36
  chips that make everything run. And finally,
• 1:39
  the just massive finance needed to actually scale
• 1:41
  it all up. It's like the foundational layers,
• 1:43
  the stuff you don't always see, but everything
• 1:45
  else relies on. Precisely. But before we dive
• 1:48
  into our Morningstar report, let's thank our
• 1:51
  first sponsor. Old Glory. Ah, yes, Old Glory.
• 1:54
  They're that great, iconic store for music and
• 1:57
  sports fan merchandise. Really cool stuff. Totally.
• 1:59
  And if you're already thinking about holiday
• 2:01
  gifts, maybe for that techie friend who seems
• 2:04
  to have everything, Old Glory is a fantastic
• 2:06
  place to look. Good idea. And you can use the
• 2:08
  promo code SNARFL. That's S -N -A -R -F -U -L
• 2:12
  at checkout to get 15 % off your entire purchase.
• 2:14
  Find all their great gear at OldGlory .com. OK,
• 2:17
  let's get into the Morningstar report. First
• 2:20
  up, something genuinely mind -bending from Google
• 2:24
  DeepMind. AI isn't just predicting things anymore,
• 2:27
  it's actually discovering and proving new mathematical
• 2:30
  truths. Yeah, this one really jumped out. We're
• 2:33
  kind of used to AI helping with, you know, writing
• 2:35
  emails or making pictures, but this feels like
• 2:37
  a huge leap, formal proof. How did they even
• 2:41
  create something to do that? Well, the system's
• 2:44
  called AlphaVolve, and what's really interesting
• 2:46
  is its approach. It's... Described as a generic
• 2:50
  evolutionary coding agent think Darwin, but for
• 2:53
  code, okay evolutionary how? It uses a large
• 2:57
  language model, an LLM, the tech behind things
• 2:59
  like ChatGPT as the idea generator. So it spits
• 3:02
  out tons of possible code solutions to a math
• 3:05
  problem. Right. But millions of ideas. Yeah.
• 3:07
  And how does it know which ones are any good?
• 3:09
  Does a human have to check them all? No, and
• 3:10
  that's the clever part. It has an evaluator.
• 3:12
  It's basically a specific mathematical fitness
• 3:14
  function. It rigorously tests all those generated
• 3:17
  code solutions against the problem's criteria.
• 3:19
  Oh, I see. So it's this closed loop. Proposed
• 3:22
  test refined proposed test refined all happening
• 3:25
  incredibly fast way faster than any human team
• 3:28
  could manage in the results Were they actually
• 3:31
  sick? They really were yeah, they tested alphavolve
• 3:33
  on 67 different math problems Geometry combinatorics
• 3:38
  number theory a real mix and get this It didn't
• 3:41
  just rediscover the best known solutions in most
• 3:44
  cases. It actually found improved solutions in
• 3:47
  about 20 of those cases. New, better ways to
• 3:51
  solve long -standing problems. Wow. Okay, improved
• 3:53
  solution in abstract math. Can you give us an
• 3:58
  analogy? What does that actually mean? Okay,
• 3:59
  think of something like the finite field kakaia
• 4:01
  conjecture It's sort of a puzzle about efficiency
• 4:04
  like packing things into abstract mathematical
• 4:07
  spaces, right? Imagine a huge high dimensional
• 4:10
  chessboard The question is something like what's
• 4:12
  the absolute smallest number of squares you need
• 4:14
  to mark so that every possible diagonal line
• 4:17
  crosses at least one mark square Okay, I think
• 4:19
  I follow AlphaVolve found a new construction
• 4:22
  for this. Essentially, a more efficient way to
• 4:24
  place those marked squares, a smaller set than
• 4:27
  mathematicians knew was possible before. It found
• 4:30
  a shortcut. That is an aha moment. Yeah. What
• 4:33
  really struck me from the reports was that this
• 4:35
  wasn't just one AI working alone, right? Absolutely
• 4:37
  not. And this really points to the future of
• 4:40
  AI research, I think. It's about specialized
• 4:42
  AI collaboration. AlphaEvolve generated the core
• 4:46
  idea, the potential solution. Then they used
• 4:49
  other specialized AI agents, things like Gemini
• 4:52
  DeepThink to actually verify the mathematical
• 4:55
  logic, and another one called AlphaProof to formally
• 4:58
  write down the proof in a specific computer language
• 5:00
  called Lean. Hold on, one AI verifying another
• 5:03
  AI's logic? How reliable is that verification?
• 5:07
  Can we trust that? That's a huge engineering
• 5:09
  challenge, right? But it shows why developing
• 5:12
  these highly specialized, rigorous AI systems
• 5:14
  just for formal proof is becoming so critical.
• 5:17
  The whole process, by the way, also involved
• 5:19
  leading human mathematicians like Taron Stowe.
• 5:21
  So it's AI augmenting human experts. Exactly.
• 5:25
  It's supercharging human intuition, letting us
• 5:27
  refine these mathematical boundaries, often by
• 5:30
  small amounts. But in incredibly complex, high
• 5:32
  dimensional problems, those small amounts can
• 5:35
  be really meaningful. OK, but why should You
• 5:37
  know, the average person care about shaving a
• 5:41
  tiny percentage point off some abstract math
• 5:44
  grid puzzle. Because that fundamental math underpins
• 5:47
  so much technology. Faster innovation there means
• 5:50
  faster real world innovation. Think about cryptography
• 5:53
  that secures your bank transfers, or the compression
• 5:56
  algorithms that let you stream this very podcast.
• 5:59
  Ah, okay. It all relies on that core math. Precisely.
• 6:02
  Better math, faster algorithms, better tech,
• 6:05
  and, predictably, the money is already following.
• 6:08
  There's a startup called Hiverge, founded by
• 6:09
  some of the DeepMind scientists who worked on
• 6:11
  this. They just raised $5 million to basically
• 6:14
  commercialize this idea of automated algorithm
• 6:16
  design. They call their platform the Hive. Right,
• 6:19
  I saw that, and their CEO had that great line
• 6:21
  saying their results are not just based on vibes.
• 6:25
  Yeah, we could probably use less vibe -based
• 6:27
  coding in the world. Definitely. Okay, so from
• 6:30
  the very abstract world of mathematical proofs,
• 6:34
  let's pivot hard to the very physical, very geopolitical
• 6:38
  world of the chip supply chain, specifically
• 6:41
  this next period crisis. Yeah, this was a big
• 6:44
  one. This conflict between the Netherlands and
• 6:46
  China over Nexperia, which is Chinese -owned
• 6:49
  but based in the Netherlands, it really did threaten
• 6:52
  to grind global manufacturing to a halt for a
• 6:55
  bit. What was the core tension there? Well, the
• 6:57
  Dutch government basically seized control of
• 6:59
  Nexperia, citing concerns about its governance.
• 7:03
  Beijing's reaction was immediate and pretty severe.
• 7:07
  They blocked all exports of Nexperia products
• 7:09
  from China. And that's a choke point because
• 7:11
  apparently 70 % of the chips and experience makes
• 7:14
  in the Netherlands have to be sent to China first
• 7:15
  for packaging before they can be shipped out
• 7:17
  globally. So bam, instant crisis. But there's
• 7:20
  been a breakthrough now, right? Things are easing
• 7:21
  up. Yes, that's the latest. China is now allowing
• 7:24
  an experience to export chips again, easing that
• 7:26
  ban. And in response, the Netherlands is apparently
• 7:29
  prepared to suspend its government control measure.
• 7:32
  And this happened after that trade truce between
• 7:35
  President Trump and President Xi Jinping in South
• 7:38
  Korea. Seems like it was directly tied to that
• 7:40
  summit. Yes, a diplomatic pressure release valve.
• 7:43
  Okay, but does this really fix the underlying
• 7:46
  governance issues the Dutch had? Or is this just,
• 7:49
  you know, kicking the can down the road to keep
• 7:51
  the car factories running? That's the million
• 7:53
  -dollar question, isn't it? It feels much more
• 7:55
  like the latter. It seems driven entirely by
• 7:57
  the immediate, potentially devastating impact
• 8:00
  this had, especially on the auto industry. Right.
• 8:03
  Remind us why an Xperia is so critical. They
• 8:06
  don't make the giant CPUs or GPUs we usually
• 8:08
  talk about, do they? No, nothing like that. They
• 8:10
  make the absolute basics. We're talking diodes,
• 8:13
  transistors, simple power control chips, stuff
• 8:16
  that costs pennies. But they're everywhere. Literally
• 8:18
  everywhere, your phone charger, your TV remote,
• 8:21
  but also critical systems in cars like airbag
• 8:24
  controllers, braking systems. Almost any electronic
• 8:27
  device needs these fundamental components. Which
• 8:30
  explains why the auto industry went into panic
• 8:32
  mode almost immediately. They were already dealing
• 8:35
  with tight inventories post pandemic. Exactly.
• 8:38
  The impact was concrete and fast. Honda had to
• 8:41
  suspend production at a plant in Mexico. Stellantis
• 8:44
  apparently set up a literal war room just to
• 8:46
  monitor the Nixperia situation day by day. Wow.
• 8:49
  So yeah, this resolution averted immediate shutdowns,
• 8:52
  which is good news. But it really starkly highlights
• 8:55
  how fragile these global supply chains still
• 8:58
  are. The smallest, cheapest component can suddenly
• 9:00
  become a massive geopolitical lever and hold
• 9:03
  trillion -dollar industries hostage. A stark
• 9:06
  reminder indeed. Crisis averted for now, but
• 9:08
  the vulnerability is still there. Okay, let's
• 9:11
  move to our third story. This one involves the...
• 9:14
  just staggering scale of AI investment and a
• 9:18
  very public, maybe accidental comment from OpenAI
• 9:21
  CFO Sarah Friar. Yeah, this is a fascinating
• 9:24
  example of how careful executives need to be
• 9:26
  with their language right now, especially when
• 9:28
  talking big money. Friar was on stage at the
• 9:31
  Wall Street Journal TechLive event, and she was
• 9:33
  talking about financing OpenAI's absolutely massive
• 9:35
  infrastructure needs. They've made commitments
• 9:37
  recently totaling, get this, over $1 .4 trillion.
• 9:41
  $1 .4 trillion. That number itself just demands
• 9:45
  scrutiny. How on earth do you finance that kind
• 9:48
  of build out? Well, that was the question and
• 9:50
  friar mentioned needing a whole ecosystem Banks
• 9:53
  private equity and then she said maybe even Governmental
• 9:58
  the backstop the guarantee that allows the financing
• 10:01
  to happen the backstop. That's the key word Exactly.
• 10:04
  The moderator immediately picked up on it asking
• 10:07
  for clarification some federal backstop for chip
• 10:10
  investment and she said she confirmed it exactly
• 10:13
  that single word backstop set well pretty immediate
• 10:16
  ripples because backstop implies government support
• 10:20
  government risk reduction it suggests the scale
• 10:22
  of AI investment might be becoming too big to
• 10:25
  fail yeah right like the banks in 08 that's precisely
• 10:28
  how it was interpreted by many The reaction was
• 10:30
  sharp enough that later that same day, Fryer
• 10:32
  went on LinkedIn to issue a correction. A walk
• 10:35
  back. What did she say? She wrote that her use
• 10:37
  of the word backstop had, quote, muddied the
• 10:39
  point. And she insisted OpenAI is not seeking
• 10:42
  a government backstop or guarantee. Okay. But
• 10:45
  was it just a poor choice of words, a genuine
• 10:48
  gaffe under pressure? Or, you know, could it
• 10:51
  have been a subtle way to float the idea given
• 10:53
  how astronomical that $1 .4 trillion figure is?
• 10:56
  Yeah. People are already questioning if that
• 10:58
  level of spending is even viable long term. It's
• 11:00
  impossible to know her intent for sure. Right.
• 11:02
  But regardless, the slip up tap directly into
• 11:05
  that widespread concern is this AI build out
• 11:08
  becoming so enormous, so potentially unstable
• 11:12
  that it might evade eventually near public support.
• 11:15
  Fryer tried to pivot the message saying she just
• 11:17
  meant that both the private sector and government
• 11:19
  need to work together to grow the US's industrial
• 11:22
  capacity and that AI should be viewed as a national
• 11:25
  strategic asset. Which raises its own big questions
• 11:28
  about what a national asset even means today.
• 11:30
  Is it roads and bridges or is it massive AI compute
• 11:33
  power? It's a fundamental question. And the fact
• 11:35
  that one word potentially misused can spark such
• 11:39
  intense global discussion about a $1 .4 trillion
• 11:41
  venture shows how incredibly sensitive and high
• 11:45
  stakes this whole intersection of AI technology
• 11:48
  and finance has become. Absolutely. Okay, before
• 11:51
  we wrap things up, let's quickly thank our other
• 11:53
  fantastic sponsor, Scony. Ah, Scony, yes. The
• 11:57
  premium golf shoe store. If you're looking for
• 11:59
  some seriously stylish and comfortable shoes
• 12:01
  out on the course, Skoni is definitely worth
• 12:04
  checking out. Totally agree. And good news for
• 12:06
  our listeners, if you use that same promo code,
• 12:09
  SNARFUL, S -N -A -R -F -U -L at checkout, you
• 12:12
  get 15 % off your entire purchase there too.
• 12:14
  Find them online at skoni .com. That's S -K -O
• 12:18
  -N -I dot com. All right. So let's quickly connect
• 12:21
  the dots on what we covered today. Three really
• 12:24
  critical pressure points in tech. Yeah, first
• 12:26
  we saw AI with Elf Evolve moving beyond just
• 12:29
  predicting things to actually discovering new
• 12:32
  mathematical structures. It shows even the absolute
• 12:34
  foundations of math are still evolving, which
• 12:36
  is kind of wild. We had an experience situation,
• 12:39
  a geopolitical showdown, over tiny essential
• 12:41
  chips that reminded us just how much leverage
• 12:44
  exists in controlling even the cheapest parts
• 12:45
  of the supply chain. Enough leverage to force
• 12:48
  a trade truce, apparently. And finally, that
• 12:50
  open AI story. the sheer financial scale of building
• 12:53
  the AI future, and how one executive's comment,
• 12:56
  maybe a slip up, maybe not, highlighted the massive
• 12:59
  skepticism and high stakes surrounding that $1
• 13:02
  .4 trillion race. It really feels like these
• 13:05
  stories show that the biggest tech news today
• 13:07
  isn't always about the latest gadget or app speed,
• 13:10
  right? Absolutely. It's often about the fundamental
• 13:12
  math, the security of manufacturing for even
• 13:15
  the smallest components, and the almost unbelievably
• 13:18
  huge, maybe fragile financing that underpins,
• 13:21
  well, pretty much everything in modern life now.
• 13:23
  We're here to help at Snarful, so reach out on
• 13:25
  our website if you have any more questions. That's
• 13:31
  a wrap on today's top stories. Remember, tech
• 13:33
  doesn't have to be complicated and we're here
• 13:36
  to help connect the dots. If you've got questions
• 13:39
  or want to learn more, visit us at snarful .com.
• 13:42
  Thanks for listening and we'll catch you next
• 13:44
  time.