lee/hpr-knowledge-base
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Initial commit: HPR Knowledge Base MCP Server

Browse Source 
- MCP server with stdio transport for local use
- Search episodes, transcripts, hosts, and series
- 4,511 episodes with metadata and transcripts
- Data loader with in-memory JSON storage

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Lee Hanken
2025-10-26 10:54:13 +00:00
commit 7c8efd2228
4494 changed files with 1705541 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

166
hpr_transcripts/hpr4476.txt Normal file
View File

@@ -0,0 +1,166 @@
Episode: 4476
Title: HPR4476: Does AI cause brain damage?
Source: https://hub.hackerpublicradio.org/ccdn.php?filename=/eps/hpr4476/hpr4476.mp3
Transcribed: 2025-10-26 01:04:09
---
This is Hacker Public Radio Episode 4476 from Monday 29 September 2025.
Today's show is entitled, Does AI Cause Brain Damage?
It is hosted by Troubleder Coaster and is about 12 minutes long.
It carries an explicit flag.
The summary is, going through the findings of an MIT study about how AI will make your brain
rot away or will it.
Hello knowledge seekers, doose strollers and heroic procrastinators.
Welcome aboard the Troubleder Coaster.
The only ride where the safety bars is smug green and the drops are powered by sheer
cognitive dissonance.
Today we're asking a radical question.
Does thinking still matter in 2025?
Spoiler MIT just ran an experiment suggesting the answer is, yeah, kinda.
And the internet promptly tried to cancel them for H speech against convenience.
Strap in sarcasm, heightened learning, peer review pending.
Picture, 54 volunteers each wearing an EEG swim cap that makes them look like budget marvel villains.
These brave souls were told to ride three short essays under three different conditions.
The first team ride their essay using brain only.
Raw, neural, horsepower, no training wheels.
The second team were allowed to use classical search engines.
Let's call them, let me google that for you crowd.
The third team could use Chatchy PT to help them ride their essays.
After three rounds of writing an essay, 18 volunteers earned a bonus level.
The LLM group had their shiny toy yanked away while the brain only warriors finally got to taste the forbidden fruit.
So EEGs scanned the brains and repeated some whizz back to evaluate the originality of the essays.
Teachers graded those essays and so did the NEI judge, since they already paid for the opening account anyway.
The results? Wow, you wouldn't have expected it.
The brain only team lit up the neural fireworks display, higher activity on every possible level.
More lights than the server rack of Jeff Geerling.
Search users still showed respectable sparks and the LLM group.
Let's politely say the range can't look like a single birthday candle nobody bothered to blow out.
Ownership scores tanked and the test group barely remembered the quotes they themselves well actually their AI had typed.
And their essays, clones, all the same structure, all the same topics, ba ba ba.
But wait, there's nuance.
The researchers did not scream AI will melt your brain.
No, they coined the calmer term, they called it cognitive death.
Offload too much thinking today and you pay interest tomorrow, less in engagement or weaker recall.
And the vague sensation, your originality is slowly being garnished.
Before we torch every chatbot, note the studies fine print, only 54 participants.
And in the switch round, only 18, this is an exploratory postcard from the frontier, not the final gospel.
So hold the pitchforks or at least keep the tags on for easy returns.
Now comes the juicy part.
When the LLM loyalists had their crutch removed, e.e. engagement tanked faster than my motivation on a Monday.
Meanwhile, brain first folks who finally got to use AI still remembered what they had written two weeks ago,
kept the neural party going and basically showed that adding tools after you build muscles is fine.
Replacing muscles with tools less fine.
It's almost as if practice matters.
Now isn't that a crazy concept? Travel back to your childhood with me.
Remember those memorizing those tables and multiplication, hitting every minute of it.
And if you finally learned how to bend your head around it and you could drill them like a sergeant,
one, two, three, four, they gave you a calculator and feeling like four welding mealier.
Key detail.
You learned before you offloaded.
Nobody said skip edition kiddo, the TI-84s got your back.
Nope, you had to learn them anyway because it helps you.
Translating this to 2025, build cognitive strength first and then unleash the silicon sidekick.
So it's clear it's not a new insight, scuffles, beat shortcuts.
Need a real more example here in Belgium, July.
Medical entrance exam.
A bunch of applicants, slipchatchypt answers through the digital loophole.
And the headline writers rejoice.
Parents panic and somewhere a sergeant sighs because yet again people conflate using tools with being tools.
Authorities investigated.
Found a handful, quite a few bad apples, and reminded everyone that yes, future doctors should know where the heart is.
Without crowdsourcing the question, revolutionary stuff.
The Dutch professor, filling her months nails it, which he says,
relying on an LLM to do your homework is like rolling a fork lift into the gym to bench press for you.
Sure, the weights go up, but your biceps stays as squishy as a half heated marshmallow.
No pain, no gain, the same goes for the brain.
Both demand resistance to grow.
And if you're thinking, whatever, I'll just multitask my way into greatness.
Decades of studies on digital multitasking would love a word.
Heavy jugglers of screens, often sport, weaker working memory, attention spans or even a lack of self-discipline.
There are even structural brain differences.
Turns out your brief frontal cortex is not a fan of 15-brothered trap tabs all screaming.
Click me, click me, click me, click me.
Again, who would have guessed?
It's established.
No toys for the kiddies, but my brain, it's all grown up.
So I don't have to worry about all this chisel.
I can play with Claude and chat with Petey to my heart's content because we all know brain degradation isn't a thing.
There's one other cute elephant quietly sleeping in the room, hoping this drawer coaster will just steer away from it.
The elephant is called the hype cycle.
It's where every new technology gets pumped into the market as the next coming of Jesus that will save us all.
Startups experiment with the new tech and savvy investors pump and hype in the hope to make some money out of it.
The big companies like Microsoft and Google will ingest successful start-ups making it their own.
A few other successful start-ups will manage to become their own big beast.
You thought you were paying for your Chachi beast description?
Think again, you're merely paying for the lipstick.
Smoking up their companies, swaying their hips for the Bitcoin barons and digital landlords.
And to be able to do that, they need to show off their big customer base.
Promising big gains behind the curtains.
So a bunch of investors took the bait.
Millions even billions are rolling in.
And after a while though, the investors want their money back.
Ideally, with some extras as that's what they were promised.
And this, my friend, is where the prices will go up and features go away.
Cory Doctorow nicely coined it in certification, but that's for a whole other podcast.
Maybe one you'll make because, come on, let's be honest.
You don't want another one of these insulting episodes on our public radio, right?
Just one last thought on this.
The secret sauce is called open sauce.
But don't tell it forward or maybe do.
For now, let's just roll over and accept that other limbs are here to stay.
And maybe can be useful too.
I challenge you!
Yes, you, with your tiny earbuds or your 20th century oversized headset,
to run a two day experiment.
The next time you need to create some content, maybe an essay.
Maybe your love letter.
Oh, please.
Don't let chat GPT write your love letters.
Maybe that nasty resignation mill you always dreamt of writing.
Or maybe the outline for an episode of hacker public radio.
Step one.
Use your grey matter to create the outline.
Nothing else.
No tech.
Just a piece of paper.
And steer away from the siren song of autocomplete and write it yourself.
Then, if it's done, invite your favorite Allen to critique, expand or dunk on your ideas.
If the Allen is done stroking your ego and along the way also triggered some ideas,
send your artificial voice for a nappy again.
Finish with a personal rewrite.
And 48 hours later, test what you remember out of the top of your head.
What do you remember yourself?
If your mind's a blank canvas, congratulations.
You've just paid the minimum on your cognitive debt credit card.
Time to refinance it with more effort.
So, should we ditch the bit chatbot right away?
Short answer.
Only if you're also eating your calculator in doom and doom.
Long answer.
Treat AI as a scaffolding.
Keep thinking in the driver's seat.
And remember, that speed means nothing if you swap accuracy for autopilot.
foundational fluency first.
Acceleration second.
That's how calculators empowered math rather than using it.
And that's how elevators can make us smarter instead of dollar.
Alright riders, the troller coaster is pulling back into the station.
Brains intact, I hope.
Key souvenirs, cognitive debt, is real but manageable.
Effort precedes offloading and a pinch of sarcasm spices things up.
So, you don't agree?
Am I too nice to the AI overlords?
Or did I hurt the feelings of your precious little psychophantic chatbody?
Ha ha ha ha, great!
Flip out the data that proves I'm a full of digital hot air.
Record your own podcast, season it with a bit of spicy sarcasm.
Don't spare me, I can take it, I like it hot.
But if you're more of the sweet type, feel free to sugarcoat it.
Until then, keep your neurons firing and have your AI on a leash.
Preferably a retractable one.
You have been listening to Hacker Public Radio at Hacker Public Radio does work.
Today's show was contributed by a HPR listener like yourself.
If you ever thought of recording a podcast, click on our contribute link to find out how easy it really is.
Hosting for HPR has been kindly provided by an honesthost.com, the Internet Archive and our Sync.net.
On the Sadois status, today's show is released under Creative Commons Attribution 4.0 International License.