Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
what up nerds I'm
0:02
Jared and this is
0:04
change log news for
0:07
the week of Monday
0:09
March 10th 2025 vibe
0:11
coding is the valley's
0:13
buzz word du jour
0:15
for the uninitiated it's
0:17
like pair programming with
0:19
an AI Where it
0:21
writes all the code
0:23
and your only job
0:25
is to make sure
0:27
it doesn't harsh your
0:29
vibe, dude. Then, when
0:31
it's finished, you post
0:33
a demo to social
0:35
media, declare software engineering dead,
0:37
throw the barely usable final
0:40
product away, and code up
0:42
some more vibes. Okay, let's
0:44
get in to this week's
0:46
news. Everyone is talking about
0:49
MCP. Okay, probably not everyone
0:51
is talking about MCP. but
0:53
it's certainly a burgeoning topic
0:55
amongst the AI engineering crowd.
0:58
So I figured it's at
1:00
least worth a primer here.
1:02
MCP stands for model context
1:04
protocol which was first announced
1:06
by the anthropic team last
1:09
November. It's an open protocol
1:11
to standardize how applications provide
1:13
context to LLLM's. Quote, think
1:15
of MCP like a USBC
1:18
port for AI apps. Just
1:20
as USBC provides a standardized
1:22
way to connect your devices
1:24
to various peripherals and accessories,
1:27
MCP provides a standardized way to
1:29
connect AI models to different data
1:31
sources and tools. End quote. The
1:34
linked X thread lists out a
1:36
bunch of things people are building
1:38
and announcing around MCP. Most of
1:41
which look like demo quality wears,
1:43
but there's certainly some potential here.
1:45
If we're going to have an
1:47
agentic future. We need good ways
1:49
to equip AI agents with the
1:51
context they need to accomplish their
1:53
tasks. MCP might become the way
1:56
we achieve that, or perhaps just
1:58
a step along the way. Stop. algorithms
2:00
dictate how you think. Here's
2:02
Tom Usher. Quote, the creators
2:04
of TikTok, Instagram, etc. have
2:07
gained control over exactly what
2:09
we see. And what we
2:11
see strongly influences how we
2:13
think. They know their feeds
2:15
make us angry. They know the
2:17
negative effects on our mental health,
2:19
particularly that of teens. And they
2:22
know that they have an influence on
2:24
our opinion. With the power to shape,
2:26
what we see comes the power to
2:28
shape what we believe. Whether through
2:30
deliberate manipulation or the slow
2:32
creep of algorithmic recommendations, engagement
2:35
is fueled by outrage and
2:37
outrage breeds extremism. The result
2:39
is a feedback loop that
2:42
isolates users, reinforces beliefs, and
2:44
deprioritizes opposing viewpoint. End quote.
2:47
Being able to form our own opinions
2:49
is more important than ever. Do you
2:51
want to take back control of how
2:53
you think? At the end of
2:55
this post, Tom gives you five
2:57
things you can do. without going
2:59
cold turkey off social media all
3:01
together. Troubleshooting never goes obsolete. In
3:03
an industry where it's too easy
3:06
to invest time in skills that
3:08
don't last, I love when people
3:10
share their expertise on things that
3:12
have stood the test of time.
3:14
Here's curiosity doing just that. Quote,
3:16
realizing that I spend more time troubleshooting
3:18
than I do building or doing
3:20
and that the skill of troubleshooting
3:22
can be honed separately from the
3:25
domain it's applied to. I decided
3:27
to try to figure out how
3:29
to improve my troubleshooting skills, and
3:31
as a result, my effectiveness in
3:33
multiple domains. The way I do
3:35
it, troubleshooting mostly boils down to
3:37
scratching my head, googling the error
3:40
message, and thinking up and testing
3:42
hypotheses to narrow the search space.
3:44
But I frequently catch myself making
3:46
errors I have made before, so
3:48
here's what I try to remember when
3:50
I'm troubleshooting to keep myself on track
3:53
and avoid dead ends. That last
3:55
paragraph is an extreme simplification.
3:57
This essay is brimming with
3:59
specific. High quality advice. Check it
4:01
out, link in the newsletter.
4:03
It's now time for sponsored news.
4:05
New temporal capabilities from Replay
4:08
2025. The biggest event of the
4:10
year for Temporal just wrapped
4:12
up. Replay 2025 in London is
4:14
where teams learn about evolving
4:16
legacy systems, modernizing infrastructure, and building
4:18
scalable, durable software to power
4:20
today's top businesses. Here is what
4:22
they announced. Temporal is open
4:25
source, so you can easily self
4:27
host it. But many teams
4:29
are operating at scale and they
4:31
rely on Temporal Cloud to
4:33
give them an easier, faster, more
4:35
scalable way to run Temporal.
4:37
You can now automate migrations with
4:39
zero downtime. The new tooling
4:42
is in pre -release and they're
4:44
seeking early users. High availability can
4:46
now offer another nine to
4:48
multiple region replication and same region
4:50
replication allowing you to asynchronously
4:52
replicate your workflows to a namespace
4:54
in a secondary or same
4:56
region and will automatically fail over
4:59
if necessary to keep your
5:01
apps online with a 99 .99
5:03
SLA. They also released new and
5:05
updated capabilities to improve your
5:07
DX with Temporal. Temporal Ruby SDK
5:09
is now available in pre -release
5:11
and is that full feature
5:13
parity with the other Temporal SDKs.
5:16
You can now use Ruby
5:18
to write workflows and activities. And
5:20
Temporal Nexus is now generally
5:22
available. Nexus lets you connect Temporal
5:24
applications across and within isolated
5:26
namespaces. Check the link in the
5:28
newsletter for all the announcements
5:30
and thank you to Temporal for
5:33
sponsoring Change Log News. Write
5:35
blogs so LLMs have something
5:37
to read. Why create content in
5:39
the age of AI slop?
5:41
Nikola Duzza has a dystopian answer
5:44
to that question. Quote, LLMs
5:46
are getting better and better but
5:48
they all need some kind
5:50
of input to be trained on.
5:52
And that's where we, content
5:54
creators, come in. The world needs
5:56
human touch for now. The
5:58
AI needs some human touch. The
6:00
LLMs. LMs need to be trained on the good stuff.
6:02
That's why we need to keep
6:04
writing, recording, and creating." This is giving
6:06
me strong matrix vibes. Spoiler alert.
6:09
You know the matrix where the
6:11
machines only reason to keep us
6:13
alive is to harvest us for
6:15
energy. So they create a digital
6:17
world to keep our minds busy
6:19
while they feed on our bodies.
6:21
But maybe I'm the only one
6:23
getting matrix vibes. I do agree
6:25
with Nicola on this point. In
6:27
the sea of generated content, the
6:29
custom, handmade, locally produced content will
6:32
always stand out. It's the human touch
6:34
that will make the difference. Furthermore,
6:36
blogs and writing will highly unlikely
6:38
die. To write is a way
6:40
of getting your thoughts calm, organized
6:42
and composed. Certain humans will always
6:44
need to write, no matter how
6:46
easy it is to generate content
6:48
with AI. In that sense, blogs
6:50
with human touch will always have
6:52
a place. End quote. Always is
6:54
a strong word, but I do
6:56
believe there will be a market
6:58
for human crafted content for the
7:00
foreseeable future, just like there's still
7:02
a market for handcrafted goods hundreds
7:04
of years after the Industrial Revolution.
7:06
Should managers still code? James Staneer
7:08
takes a crack at this age-old
7:10
question. Quote, the short answer is
7:12
that it depends exactly on what
7:14
you mean by coding. I think
7:16
that there is a big difference
7:19
between being in the code and
7:21
writing code. All managers should be
7:23
in the code, but not all managers
7:25
should be writing code." James spends some
7:27
time digging into the nuances of
7:29
the question and how exactly he
7:31
wants his engineering managers to be
7:33
in terms of their relationship with
7:36
the code base. Should they be
7:38
able to write code? Should they be
7:40
able to do code reviews? Should they
7:42
be able to debug and triage
7:44
production issues? Stuff like that. It's
7:46
a good one. Check it out
7:48
in the newsletter. That's the news
7:51
for now, but also scan that
7:53
companion newsletter for even more links
7:55
worth clicking on, such as building
7:57
websites with lots of little HDML
7:59
pages. It is as if
8:01
you were on your phone.
8:04
Her crypt, ferociously secure, positively
8:06
adorable. And, keep your calendar
8:08
in a plain text file.
8:10
Sign up today for $0
8:13
at change law.com/news. This is
8:15
episode number 135, so that
8:17
means it's time, once again,
8:20
for some change log, plus
8:22
plus shout-outs. Shout out to
8:24
our newest members. Yaraslaw, FELIXIX.
8:26
Felix, KED. Alex, G, Matt,
8:29
T, and Scott, S. We appreciate
8:31
you for supporting our work with
8:33
your hard-earned cash. If Change Log
8:35
Plus Plus Plus is news to
8:38
you, that's our membership program. You
8:40
can join to dish the ads.
8:42
Get in on bonus content. Receive
8:44
a free sticker pack in the
8:46
mail. And get shout-outs, like the
8:48
ones you just heard. Change Log
8:50
Plus Plus Plus! It's better. Have
8:52
a great week. Leave us a
8:55
five-star review, if you dig the
8:57
show. And I'll talk to you
8:59
again. real soon.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More