1
00:00:05,390 --> 00:00:08,910
Jennifer Reif: You are listening to the
Breaktime Tech Talks podcast, a bite-sized
2
00:00:08,950 --> 00:00:13,489
tech podcast for busy developers where
we'll briefly cover technical topics, news
3
00:00:13,489 --> 00:00:15,749
snippets, and more in short time blocks.
4
00:00:16,230 --> 00:00:19,940
I'm your host, Jennifer Reif, an
avid developer and problem solver
5
00:00:20,149 --> 00:00:24,100
with special interest in data,
learning, and all things technology.
6
00:00:24,919 --> 00:00:28,264
This week I had a Java User
Group tour in the Northeast US.
7
00:00:28,595 --> 00:00:31,791
It was incredibly exciting and
engaging, and I came back ready to
8
00:00:31,791 --> 00:00:33,628
use that tailwind to my advantage.
9
00:00:34,076 --> 00:00:37,762
I'm still trying to catch up on a
few things at home as well, and maybe
10
00:00:37,763 --> 00:00:41,430
recover a bit from all the travel,
but lots of things to explore.
11
00:00:41,786 --> 00:00:45,582
There was also a huge release in the
tech industry for a new type of model,
12
00:00:45,596 --> 00:00:48,775
which I will give my perspective
on and what I have found so far.
13
00:00:49,124 --> 00:00:52,718
We have a packed episode with lots
of updates today, so let's jump in.
14
00:00:53,532 --> 00:00:56,648
First of all, I want to give a
huge shout-out to the Northeast
15
00:00:56,648 --> 00:00:58,355
U.S. meetup groups from this week.
16
00:00:58,762 --> 00:01:02,314
Every group ran really smoothly,
provided an excellent experience,
17
00:01:02,335 --> 00:01:05,032
and had an audience full of
inquisitive technologists.
18
00:01:05,524 --> 00:01:08,690
There were lots of really great
questions, I had high engagement, and
19
00:01:08,692 --> 00:01:12,093
people who were genuinely interested
in learning and improving their
20
00:01:12,093 --> 00:01:15,329
technical expertise and the business
systems that they were working on.
21
00:01:16,002 --> 00:01:19,876
The first group that I spoke was at
TriJUG, which is in Raleigh, North
22
00:01:19,877 --> 00:01:24,100
Carolina, and then I moved from there
to Charlotte JUG, which is in Charlotte,
23
00:01:24,101 --> 00:01:29,098
North Carolina, the next night, and
then combined an event for the Richmond
24
00:01:29,140 --> 00:01:33,678
JUG and the GraphRVA group, which both
of those are in Richmond, Virginia.
25
00:01:33,679 --> 00:01:37,227
And I did that on the final
and third night of my travels.
26
00:01:37,958 --> 00:01:42,582
After every event, I walked away in
really high spirits, so definitely some
27
00:01:42,583 --> 00:01:45,938
kudos to the organizers and everyone
involved, and all the attendees who
28
00:01:45,939 --> 00:01:47,679
showed up on each and every night.
29
00:01:48,019 --> 00:01:51,793
If you are in any of those regions for
the groups, I highly recommend them.
30
00:01:51,945 --> 00:01:52,727
It was excellent.
31
00:01:53,514 --> 00:01:56,898
Now, we also have some exciting
things going on at Neo4j.
32
00:01:56,923 --> 00:02:00,092
The NODES 2026 agenda is now live.
33
00:02:00,369 --> 00:02:05,558
So NODES, if you are not familiar,
is Neo4j's online developer-focused
34
00:02:05,558 --> 00:02:07,226
event that runs every fall.
35
00:02:07,560 --> 00:02:11,164
We also have an AI-focused
rendition and a shorter rendition
36
00:02:11,164 --> 00:02:12,422
of this that runs in the spring.
37
00:02:12,618 --> 00:02:16,021
But this is our main key event
that runs every fall for
38
00:02:16,068 --> 00:02:17,489
online developer audiences.
39
00:02:17,771 --> 00:02:22,007
We just had the agenda go live, which
means that all of the schedule of the
40
00:02:22,020 --> 00:02:25,476
sessions, who all is speaking, what
they're talking about, the session
41
00:02:25,493 --> 00:02:29,505
abstracts, and all of the details are
now published and available on the site.
42
00:02:29,752 --> 00:02:31,916
This year we have all regular sessions.
43
00:02:31,916 --> 00:02:33,181
There aren't any lightning talks.
44
00:02:33,184 --> 00:02:39,276
So if you've been around before and
done the NODES event before, previously
45
00:02:39,277 --> 00:02:44,243
we had some longer, more in-depth
technical sessions, and then we had
46
00:02:44,245 --> 00:02:46,406
very short lightning talk sessions.
47
00:02:46,625 --> 00:02:49,629
And this year, we completely took
out the lightning talk sessions,
48
00:02:49,813 --> 00:02:53,335
and focused entirely on the
longer, more technical sessions.
49
00:02:53,548 --> 00:02:56,349
Each topic will get a better
dive in this type of format.
50
00:02:56,557 --> 00:02:59,884
But definitely reach out to us and let us
know your feedback after the event's over.
51
00:03:00,590 --> 00:03:06,132
Now, NODES 2026 will happen on November
12th of this year, and then there
52
00:03:06,132 --> 00:03:10,694
are also some pre-NODES workshops,
and they are called Road to Nodes.
53
00:03:10,975 --> 00:03:13,612
Those are available in the weeks
that are leading up to the event.
54
00:03:13,627 --> 00:03:17,193
I believe they start in the next
week, week and a half or so.
55
00:03:17,512 --> 00:03:19,771
I believe the first one might
be on October 1st, but I would
56
00:03:19,771 --> 00:03:21,397
need to check the details there.
57
00:03:21,666 --> 00:03:25,836
There are seven of them, and I will
include the link in the show notes to
58
00:03:25,836 --> 00:03:28,798
register for those if you're interested,
or at least just to check them out and
59
00:03:28,799 --> 00:03:30,299
see what all is being covered this year.
60
00:03:31,485 --> 00:03:35,081
Also, another big exciting
milestone got hit.
61
00:03:35,100 --> 00:03:37,558
All of my chapters for the
Java book are submitted.
62
00:03:37,981 --> 00:03:39,160
Now, I'm not completely done.
63
00:03:39,160 --> 00:03:43,710
I still have some editing and probably
last minute fine-tuning publishing work
64
00:03:43,734 --> 00:03:45,147
that will need to be done beforehand.
65
00:03:45,654 --> 00:03:49,124
I don't know a an official publish date
yet on this, but I will keep you informed.
66
00:03:49,332 --> 00:03:52,638
But at least the bulk of my
work on this is now complete.
67
00:03:53,380 --> 00:03:56,992
I still do have to go through a few
of the little tidbits, and I am really
68
00:03:56,992 --> 00:04:00,796
excited to start prepping for some
upcoming commitments now that I have at
69
00:04:00,796 --> 00:04:04,070
least most of the book work completed.
70
00:04:04,538 --> 00:04:06,477
I also have lots of things upcoming.
71
00:04:06,513 --> 00:04:11,052
There's some new projects on
benchmarking your own RAG pipeline,
72
00:04:11,292 --> 00:04:14,924
some multi-agent application things
I want to get into, and lots more.
73
00:04:15,137 --> 00:04:17,922
For instance, a couple of things that
surfaced from the meetup groups this
74
00:04:17,923 --> 00:04:22,349
week is there were lots of questions
on MCP security and some access
75
00:04:22,349 --> 00:04:26,904
restrictions, and how all of that
actually works in practice, which
76
00:04:26,934 --> 00:04:28,715
I haven't dug a whole lot into yet.
77
00:04:28,951 --> 00:04:30,127
So I definitely wanna explore that.
78
00:04:30,153 --> 00:04:33,569
And then other questions surfaced
on guardrails and agentic
79
00:04:33,570 --> 00:04:35,330
security, and controls around that.
80
00:04:35,598 --> 00:04:39,678
Of course, I'll talk in a second
about a new tech release that may help
81
00:04:39,679 --> 00:04:43,140
with some of those things, but again,
just something I need to dig into.
82
00:04:44,094 --> 00:04:47,900
Now, other things that are upcoming
for me, I've got some upcoming events
83
00:04:47,905 --> 00:04:51,803
that I will be working on with all of
these topics that I just discussed, and
84
00:04:51,804 --> 00:04:54,064
then also some new tech exploration.
85
00:04:54,440 --> 00:04:57,980
If you have been living under a
rock, you may not have heard about
86
00:04:58,031 --> 00:05:00,454
Jev, which is a new System 1 model.
87
00:05:00,529 --> 00:05:04,208
It just released very early this
week, and I showed up to the first
88
00:05:04,210 --> 00:05:08,275
Java user group on Monday, and people
were already asking me about it.
89
00:05:08,523 --> 00:05:13,630
And I checked my Slack, my work Slack, on
Monday afternoon before I did the meetup,
90
00:05:13,654 --> 00:05:18,837
and I already had Slack messages and
people buzzing in channels on different
91
00:05:18,838 --> 00:05:22,467
project links and things that people
were doing, and all sorts of n- news
92
00:05:22,467 --> 00:05:24,101
and noteworthy things going on there.
93
00:05:24,717 --> 00:05:30,063
Now, there's lots of coverage, so if
you're in, I, I would say pretty much any
94
00:05:30,063 --> 00:05:33,503
tech channel anywhere, you're probably
gonna see something pop up about this.
95
00:05:34,056 --> 00:05:38,024
Already, there are loads of projects and
integrations that are spinning up as well.
96
00:05:38,842 --> 00:05:41,354
Let me just give you my initial take.
97
00:05:41,683 --> 00:05:45,707
Now, this is just my perspective
and just a few little tidbits and
98
00:05:45,707 --> 00:05:47,620
notes that I've come across thus far.
99
00:05:47,982 --> 00:05:50,635
haven't done a full-on deep dive just yet.
100
00:05:51,480 --> 00:05:56,547
But rather than outputting text,
the Jev model focuses on decision
101
00:05:56,548 --> 00:06:01,734
probabilities and more traditional
machine learning, or ML, calculations.
102
00:06:02,199 --> 00:06:07,174
As an example, we could potentially
substitute or change the way that query
103
00:06:07,176 --> 00:06:11,996
planning is done inside of databases,
and potentially outperform them as well.
104
00:06:12,185 --> 00:06:15,825
So for instance, right now, when a
database goes through and plans a
105
00:06:15,850 --> 00:06:19,697
query, whether that's Cypher or SQL
or something else, it will figure
106
00:06:19,698 --> 00:06:23,634
out which pathways it needs to take,
which which planners to use, which
107
00:06:23,643 --> 00:06:26,436
optimizations to tack on, and so on.
108
00:06:26,946 --> 00:06:33,516
But potentially, this model, the System
1 model in Jev, could potentially show
109
00:06:33,518 --> 00:06:38,278
the probabilities for taking different
routes and choosing those optimal
110
00:06:38,278 --> 00:06:42,010
routes based on the more traditional
machine learning calculations.
111
00:06:42,781 --> 00:06:46,984
This is also likely a good candidate
for an agent model that will make
112
00:06:46,985 --> 00:06:48,690
decisions on which tools to call.
113
00:06:49,069 --> 00:06:53,934
So as an example, this week I talked
about agents, tools, and MCP, and
114
00:06:53,997 --> 00:06:59,070
right now a chat model is being used
for the agent decision-making process.
115
00:06:59,071 --> 00:07:03,801
It will pick and choose which tool it
thinks will best complete the input
116
00:07:03,813 --> 00:07:06,810
that was provided, either best answer
the question or complete the task
117
00:07:06,811 --> 00:07:07,988
or whatever it is that's coming in.
118
00:07:08,609 --> 00:07:12,899
However, chat models are more
non-deterministic and black boxed
119
00:07:12,947 --> 00:07:15,259
than most of us in tech probably like.
120
00:07:15,523 --> 00:07:19,906
For instance, most of the time it might
go with tool number three to answer this
121
00:07:19,936 --> 00:07:24,169
particular problem, but every great once
in a while it might pick a combination of
122
00:07:24,169 --> 00:07:29,556
two and four, or it might pick tool five,
and we have no idea why it, it does that.
123
00:07:30,283 --> 00:07:34,074
But instead, let's say that we have
seven tools to choose from, and Jev
124
00:07:34,094 --> 00:07:38,633
calculates the percentage of each tool's
likelihood to solving that input, and
125
00:07:38,633 --> 00:07:43,284
then we can understand why it picked that
tool and how heavy of a percentage over
126
00:07:43,292 --> 00:07:45,927
other tools it might have optimized for.
127
00:07:46,861 --> 00:07:51,532
Which means the response here, the
answer, the chosen tool, will be
128
00:07:51,532 --> 00:07:55,218
much more consistent and reliable
for picking from a list of options.
129
00:07:55,635 --> 00:07:58,918
Now, I also had someone tell
me that Spring AI already
130
00:07:58,924 --> 00:08:00,504
has support for Jev as well.
131
00:08:00,770 --> 00:08:03,981
I have the blog post open, just haven't
gotten all the way through it yet.
132
00:08:04,197 --> 00:08:07,212
So I will leave that in the show notes
as well, but that's something that
133
00:08:07,212 --> 00:08:10,866
I want to play with probably pretty
quickly, and one of my first entry
134
00:08:10,866 --> 00:08:12,902
points into this new technology.
135
00:08:13,981 --> 00:08:14,316
All right.
136
00:08:14,331 --> 00:08:17,013
For content pieces, I have a couple.
137
00:08:17,275 --> 00:08:21,056
The first one is one that's been
open in my browser tabs for a
138
00:08:21,057 --> 00:08:24,498
while, and it's called The Startup's
Guide to Getting Started Building
139
00:08:24,509 --> 00:08:26,957
on Neo4j Aura by Pradeep Ponduri.
140
00:08:27,289 --> 00:08:32,969
Now, I had seen this blog post go
live few weeks ago and just hadn't
141
00:08:32,969 --> 00:08:35,078
gotten a chance to walk through it.
142
00:08:35,452 --> 00:08:41,274
But if you're not familiar, Neo4j
revamped its startup program in the
143
00:08:41,276 --> 00:08:45,977
last several months, I guess it is
now, and it has a relatively new
144
00:08:45,982 --> 00:08:50,256
launch with Neo4j Aura credits, which
is our cloud database as a service.
145
00:08:50,479 --> 00:08:54,443
So now startups can get access
to credits to run on Neo4j Aura.
146
00:08:55,218 --> 00:08:59,056
This particular blog post includes all
of the resources to what you need to
147
00:08:59,057 --> 00:09:03,434
get up and running, starting from using
the Neo4j CLI from a coding agent.
148
00:09:03,691 --> 00:09:05,339
So how do you provision an instance?
149
00:09:05,345 --> 00:09:06,215
How do you create it?
150
00:09:06,225 --> 00:09:10,604
How do you start working through your
model and queries and importing and so on?
151
00:09:10,877 --> 00:09:14,800
The blog post includes links to all
of the key help that you might need,
152
00:09:14,977 --> 00:09:18,839
things like free training courses,
data science and GraphRAG library
153
00:09:18,863 --> 00:09:20,948
links and resources, and so much more.
154
00:09:21,234 --> 00:09:25,395
I really loved this overview and thought
it did a really great job at pointing
155
00:09:25,396 --> 00:09:28,889
out where you might need to go for
certain things, depending on what you're
156
00:09:28,889 --> 00:09:30,461
looking for in building your startup.
157
00:09:30,709 --> 00:09:35,332
Definitely check this out if you're just
curious what Neo4j offers or has available
158
00:09:35,349 --> 00:09:39,198
in the Neo4j startup program, or if you
are interested in becoming part of it.
159
00:09:40,380 --> 00:09:43,894
Now, the other blog post I have
follows up on a bit of what I was
160
00:09:43,895 --> 00:09:45,369
talking about earlier with Jev.
161
00:09:45,705 --> 00:09:49,722
Michael Hunger at Neo4j already
published a blog post called Navigating
162
00:09:49,733 --> 00:09:51,697
a Neo4j Knowledge Graph with Jev.
163
00:09:52,141 --> 00:09:54,956
I'll include all the, the
links to these posts as well.
164
00:09:55,786 --> 00:09:59,725
Now, Michael didn't have access
to Jev, the GA rendition just yet.
165
00:09:59,758 --> 00:10:03,350
This was over, I guess, the end
of last week or over the weekend.
166
00:10:03,831 --> 00:10:07,267
But he started exploring before the
release, and there are three different
167
00:10:07,274 --> 00:10:08,944
operations that are available in Jev.
168
00:10:09,197 --> 00:10:12,375
The first one is a choice,
where the model will choose the
169
00:10:12,385 --> 00:10:13,862
best from a list of options.
170
00:10:14,302 --> 00:10:17,330
Then the other one is a noul
which is a score from zero to
171
00:10:17,330 --> 00:10:19,363
one of how true is a statement.
172
00:10:19,806 --> 00:10:24,349
The third option for operations is
a score, which is numeric scoring.
173
00:10:24,523 --> 00:10:27,202
I don't think the numeric scoring
was used specifically in this
174
00:10:27,204 --> 00:10:28,633
blog post, but the other two were.
175
00:10:29,016 --> 00:10:32,669
Michael walks through a demo for
how he constructed the sample
176
00:10:32,669 --> 00:10:36,778
code, along with some pros and
cons of his initial exploration.
177
00:10:37,051 --> 00:10:40,263
I definitely hope to be integrating
Jev and trying this out in my
178
00:10:40,264 --> 00:10:42,220
own projects very, very soon.
179
00:10:44,031 --> 00:10:46,983
This week, I was completely impressed
with the Meetup groups that I visited,
180
00:10:46,998 --> 00:10:50,632
and came back with a full plate of
energy and new things to explore.
181
00:10:51,005 --> 00:10:54,583
Tech is moving so quickly and doesn't
seem to stop, which was reinforced
182
00:10:54,583 --> 00:10:55,944
with this week's buzzing release.
183
00:10:56,329 --> 00:10:59,469
From upcoming events to new
things to explore, it is sure to
184
00:10:59,469 --> 00:11:01,108
keep us busy for weeks to come.
185
00:11:01,486 --> 00:11:03,924
As always, thanks for
listening, and happy coding.
00:00:05,390 --> 00:00:08,910
Jennifer Reif: You are listening to the
Breaktime Tech Talks podcast, a bite-sized
2
00:00:08,950 --> 00:00:13,489
tech podcast for busy developers where
we'll briefly cover technical topics, news
3
00:00:13,489 --> 00:00:15,749
snippets, and more in short time blocks.
4
00:00:16,230 --> 00:00:19,940
I'm your host, Jennifer Reif, an
avid developer and problem solver
5
00:00:20,149 --> 00:00:24,100
with special interest in data,
learning, and all things technology.
6
00:00:24,919 --> 00:00:28,264
This week I had a Java User
Group tour in the Northeast US.
7
00:00:28,595 --> 00:00:31,791
It was incredibly exciting and
engaging, and I came back ready to
8
00:00:31,791 --> 00:00:33,628
use that tailwind to my advantage.
9
00:00:34,076 --> 00:00:37,762
I'm still trying to catch up on a
few things at home as well, and maybe
10
00:00:37,763 --> 00:00:41,430
recover a bit from all the travel,
but lots of things to explore.
11
00:00:41,786 --> 00:00:45,582
There was also a huge release in the
tech industry for a new type of model,
12
00:00:45,596 --> 00:00:48,775
which I will give my perspective
on and what I have found so far.
13
00:00:49,124 --> 00:00:52,718
We have a packed episode with lots
of updates today, so let's jump in.
14
00:00:53,532 --> 00:00:56,648
First of all, I want to give a
huge shout-out to the Northeast
15
00:00:56,648 --> 00:00:58,355
U.S. meetup groups from this week.
16
00:00:58,762 --> 00:01:02,314
Every group ran really smoothly,
provided an excellent experience,
17
00:01:02,335 --> 00:01:05,032
and had an audience full of
inquisitive technologists.
18
00:01:05,524 --> 00:01:08,690
There were lots of really great
questions, I had high engagement, and
19
00:01:08,692 --> 00:01:12,093
people who were genuinely interested
in learning and improving their
20
00:01:12,093 --> 00:01:15,329
technical expertise and the business
systems that they were working on.
21
00:01:16,002 --> 00:01:19,876
The first group that I spoke was at
TriJUG, which is in Raleigh, North
22
00:01:19,877 --> 00:01:24,100
Carolina, and then I moved from there
to Charlotte JUG, which is in Charlotte,
23
00:01:24,101 --> 00:01:29,098
North Carolina, the next night, and
then combined an event for the Richmond
24
00:01:29,140 --> 00:01:33,678
JUG and the GraphRVA group, which both
of those are in Richmond, Virginia.
25
00:01:33,679 --> 00:01:37,227
And I did that on the final
and third night of my travels.
26
00:01:37,958 --> 00:01:42,582
After every event, I walked away in
really high spirits, so definitely some
27
00:01:42,583 --> 00:01:45,938
kudos to the organizers and everyone
involved, and all the attendees who
28
00:01:45,939 --> 00:01:47,679
showed up on each and every night.
29
00:01:48,019 --> 00:01:51,793
If you are in any of those regions for
the groups, I highly recommend them.
30
00:01:51,945 --> 00:01:52,727
It was excellent.
31
00:01:53,514 --> 00:01:56,898
Now, we also have some exciting
things going on at Neo4j.
32
00:01:56,923 --> 00:02:00,092
The NODES 2026 agenda is now live.
33
00:02:00,369 --> 00:02:05,558
So NODES, if you are not familiar,
is Neo4j's online developer-focused
34
00:02:05,558 --> 00:02:07,226
event that runs every fall.
35
00:02:07,560 --> 00:02:11,164
We also have an AI-focused
rendition and a shorter rendition
36
00:02:11,164 --> 00:02:12,422
of this that runs in the spring.
37
00:02:12,618 --> 00:02:16,021
But this is our main key event
that runs every fall for
38
00:02:16,068 --> 00:02:17,489
online developer audiences.
39
00:02:17,771 --> 00:02:22,007
We just had the agenda go live, which
means that all of the schedule of the
40
00:02:22,020 --> 00:02:25,476
sessions, who all is speaking, what
they're talking about, the session
41
00:02:25,493 --> 00:02:29,505
abstracts, and all of the details are
now published and available on the site.
42
00:02:29,752 --> 00:02:31,916
This year we have all regular sessions.
43
00:02:31,916 --> 00:02:33,181
There aren't any lightning talks.
44
00:02:33,184 --> 00:02:39,276
So if you've been around before and
done the NODES event before, previously
45
00:02:39,277 --> 00:02:44,243
we had some longer, more in-depth
technical sessions, and then we had
46
00:02:44,245 --> 00:02:46,406
very short lightning talk sessions.
47
00:02:46,625 --> 00:02:49,629
And this year, we completely took
out the lightning talk sessions,
48
00:02:49,813 --> 00:02:53,335
and focused entirely on the
longer, more technical sessions.
49
00:02:53,548 --> 00:02:56,349
Each topic will get a better
dive in this type of format.
50
00:02:56,557 --> 00:02:59,884
But definitely reach out to us and let us
know your feedback after the event's over.
51
00:03:00,590 --> 00:03:06,132
Now, NODES 2026 will happen on November
12th of this year, and then there
52
00:03:06,132 --> 00:03:10,694
are also some pre-NODES workshops,
and they are called Road to Nodes.
53
00:03:10,975 --> 00:03:13,612
Those are available in the weeks
that are leading up to the event.
54
00:03:13,627 --> 00:03:17,193
I believe they start in the next
week, week and a half or so.
55
00:03:17,512 --> 00:03:19,771
I believe the first one might
be on October 1st, but I would
56
00:03:19,771 --> 00:03:21,397
need to check the details there.
57
00:03:21,666 --> 00:03:25,836
There are seven of them, and I will
include the link in the show notes to
58
00:03:25,836 --> 00:03:28,798
register for those if you're interested,
or at least just to check them out and
59
00:03:28,799 --> 00:03:30,299
see what all is being covered this year.
60
00:03:31,485 --> 00:03:35,081
Also, another big exciting
milestone got hit.
61
00:03:35,100 --> 00:03:37,558
All of my chapters for the
Java book are submitted.
62
00:03:37,981 --> 00:03:39,160
Now, I'm not completely done.
63
00:03:39,160 --> 00:03:43,710
I still have some editing and probably
last minute fine-tuning publishing work
64
00:03:43,734 --> 00:03:45,147
that will need to be done beforehand.
65
00:03:45,654 --> 00:03:49,124
I don't know a an official publish date
yet on this, but I will keep you informed.
66
00:03:49,332 --> 00:03:52,638
But at least the bulk of my
work on this is now complete.
67
00:03:53,380 --> 00:03:56,992
I still do have to go through a few
of the little tidbits, and I am really
68
00:03:56,992 --> 00:04:00,796
excited to start prepping for some
upcoming commitments now that I have at
69
00:04:00,796 --> 00:04:04,070
least most of the book work completed.
70
00:04:04,538 --> 00:04:06,477
I also have lots of things upcoming.
71
00:04:06,513 --> 00:04:11,052
There's some new projects on
benchmarking your own RAG pipeline,
72
00:04:11,292 --> 00:04:14,924
some multi-agent application things
I want to get into, and lots more.
73
00:04:15,137 --> 00:04:17,922
For instance, a couple of things that
surfaced from the meetup groups this
74
00:04:17,923 --> 00:04:22,349
week is there were lots of questions
on MCP security and some access
75
00:04:22,349 --> 00:04:26,904
restrictions, and how all of that
actually works in practice, which
76
00:04:26,934 --> 00:04:28,715
I haven't dug a whole lot into yet.
77
00:04:28,951 --> 00:04:30,127
So I definitely wanna explore that.
78
00:04:30,153 --> 00:04:33,569
And then other questions surfaced
on guardrails and agentic
79
00:04:33,570 --> 00:04:35,330
security, and controls around that.
80
00:04:35,598 --> 00:04:39,678
Of course, I'll talk in a second
about a new tech release that may help
81
00:04:39,679 --> 00:04:43,140
with some of those things, but again,
just something I need to dig into.
82
00:04:44,094 --> 00:04:47,900
Now, other things that are upcoming
for me, I've got some upcoming events
83
00:04:47,905 --> 00:04:51,803
that I will be working on with all of
these topics that I just discussed, and
84
00:04:51,804 --> 00:04:54,064
then also some new tech exploration.
85
00:04:54,440 --> 00:04:57,980
If you have been living under a
rock, you may not have heard about
86
00:04:58,031 --> 00:05:00,454
Jev, which is a new System 1 model.
87
00:05:00,529 --> 00:05:04,208
It just released very early this
week, and I showed up to the first
88
00:05:04,210 --> 00:05:08,275
Java user group on Monday, and people
were already asking me about it.
89
00:05:08,523 --> 00:05:13,630
And I checked my Slack, my work Slack, on
Monday afternoon before I did the meetup,
90
00:05:13,654 --> 00:05:18,837
and I already had Slack messages and
people buzzing in channels on different
91
00:05:18,838 --> 00:05:22,467
project links and things that people
were doing, and all sorts of n- news
92
00:05:22,467 --> 00:05:24,101
and noteworthy things going on there.
93
00:05:24,717 --> 00:05:30,063
Now, there's lots of coverage, so if
you're in, I, I would say pretty much any
94
00:05:30,063 --> 00:05:33,503
tech channel anywhere, you're probably
gonna see something pop up about this.
95
00:05:34,056 --> 00:05:38,024
Already, there are loads of projects and
integrations that are spinning up as well.
96
00:05:38,842 --> 00:05:41,354
Let me just give you my initial take.
97
00:05:41,683 --> 00:05:45,707
Now, this is just my perspective
and just a few little tidbits and
98
00:05:45,707 --> 00:05:47,620
notes that I've come across thus far.
99
00:05:47,982 --> 00:05:50,635
haven't done a full-on deep dive just yet.
100
00:05:51,480 --> 00:05:56,547
But rather than outputting text,
the Jev model focuses on decision
101
00:05:56,548 --> 00:06:01,734
probabilities and more traditional
machine learning, or ML, calculations.
102
00:06:02,199 --> 00:06:07,174
As an example, we could potentially
substitute or change the way that query
103
00:06:07,176 --> 00:06:11,996
planning is done inside of databases,
and potentially outperform them as well.
104
00:06:12,185 --> 00:06:15,825
So for instance, right now, when a
database goes through and plans a
105
00:06:15,850 --> 00:06:19,697
query, whether that's Cypher or SQL
or something else, it will figure
106
00:06:19,698 --> 00:06:23,634
out which pathways it needs to take,
which which planners to use, which
107
00:06:23,643 --> 00:06:26,436
optimizations to tack on, and so on.
108
00:06:26,946 --> 00:06:33,516
But potentially, this model, the System
1 model in Jev, could potentially show
109
00:06:33,518 --> 00:06:38,278
the probabilities for taking different
routes and choosing those optimal
110
00:06:38,278 --> 00:06:42,010
routes based on the more traditional
machine learning calculations.
111
00:06:42,781 --> 00:06:46,984
This is also likely a good candidate
for an agent model that will make
112
00:06:46,985 --> 00:06:48,690
decisions on which tools to call.
113
00:06:49,069 --> 00:06:53,934
So as an example, this week I talked
about agents, tools, and MCP, and
114
00:06:53,997 --> 00:06:59,070
right now a chat model is being used
for the agent decision-making process.
115
00:06:59,071 --> 00:07:03,801
It will pick and choose which tool it
thinks will best complete the input
116
00:07:03,813 --> 00:07:06,810
that was provided, either best answer
the question or complete the task
117
00:07:06,811 --> 00:07:07,988
or whatever it is that's coming in.
118
00:07:08,609 --> 00:07:12,899
However, chat models are more
non-deterministic and black boxed
119
00:07:12,947 --> 00:07:15,259
than most of us in tech probably like.
120
00:07:15,523 --> 00:07:19,906
For instance, most of the time it might
go with tool number three to answer this
121
00:07:19,936 --> 00:07:24,169
particular problem, but every great once
in a while it might pick a combination of
122
00:07:24,169 --> 00:07:29,556
two and four, or it might pick tool five,
and we have no idea why it, it does that.
123
00:07:30,283 --> 00:07:34,074
But instead, let's say that we have
seven tools to choose from, and Jev
124
00:07:34,094 --> 00:07:38,633
calculates the percentage of each tool's
likelihood to solving that input, and
125
00:07:38,633 --> 00:07:43,284
then we can understand why it picked that
tool and how heavy of a percentage over
126
00:07:43,292 --> 00:07:45,927
other tools it might have optimized for.
127
00:07:46,861 --> 00:07:51,532
Which means the response here, the
answer, the chosen tool, will be
128
00:07:51,532 --> 00:07:55,218
much more consistent and reliable
for picking from a list of options.
129
00:07:55,635 --> 00:07:58,918
Now, I also had someone tell
me that Spring AI already
130
00:07:58,924 --> 00:08:00,504
has support for Jev as well.
131
00:08:00,770 --> 00:08:03,981
I have the blog post open, just haven't
gotten all the way through it yet.
132
00:08:04,197 --> 00:08:07,212
So I will leave that in the show notes
as well, but that's something that
133
00:08:07,212 --> 00:08:10,866
I want to play with probably pretty
quickly, and one of my first entry
134
00:08:10,866 --> 00:08:12,902
points into this new technology.
135
00:08:13,981 --> 00:08:14,316
All right.
136
00:08:14,331 --> 00:08:17,013
For content pieces, I have a couple.
137
00:08:17,275 --> 00:08:21,056
The first one is one that's been
open in my browser tabs for a
138
00:08:21,057 --> 00:08:24,498
while, and it's called The Startup's
Guide to Getting Started Building
139
00:08:24,509 --> 00:08:26,957
on Neo4j Aura by Pradeep Ponduri.
140
00:08:27,289 --> 00:08:32,969
Now, I had seen this blog post go
live few weeks ago and just hadn't
141
00:08:32,969 --> 00:08:35,078
gotten a chance to walk through it.
142
00:08:35,452 --> 00:08:41,274
But if you're not familiar, Neo4j
revamped its startup program in the
143
00:08:41,276 --> 00:08:45,977
last several months, I guess it is
now, and it has a relatively new
144
00:08:45,982 --> 00:08:50,256
launch with Neo4j Aura credits, which
is our cloud database as a service.
145
00:08:50,479 --> 00:08:54,443
So now startups can get access
to credits to run on Neo4j Aura.
146
00:08:55,218 --> 00:08:59,056
This particular blog post includes all
of the resources to what you need to
147
00:08:59,057 --> 00:09:03,434
get up and running, starting from using
the Neo4j CLI from a coding agent.
148
00:09:03,691 --> 00:09:05,339
So how do you provision an instance?
149
00:09:05,345 --> 00:09:06,215
How do you create it?
150
00:09:06,225 --> 00:09:10,604
How do you start working through your
model and queries and importing and so on?
151
00:09:10,877 --> 00:09:14,800
The blog post includes links to all
of the key help that you might need,
152
00:09:14,977 --> 00:09:18,839
things like free training courses,
data science and GraphRAG library
153
00:09:18,863 --> 00:09:20,948
links and resources, and so much more.
154
00:09:21,234 --> 00:09:25,395
I really loved this overview and thought
it did a really great job at pointing
155
00:09:25,396 --> 00:09:28,889
out where you might need to go for
certain things, depending on what you're
156
00:09:28,889 --> 00:09:30,461
looking for in building your startup.
157
00:09:30,709 --> 00:09:35,332
Definitely check this out if you're just
curious what Neo4j offers or has available
158
00:09:35,349 --> 00:09:39,198
in the Neo4j startup program, or if you
are interested in becoming part of it.
159
00:09:40,380 --> 00:09:43,894
Now, the other blog post I have
follows up on a bit of what I was
160
00:09:43,895 --> 00:09:45,369
talking about earlier with Jev.
161
00:09:45,705 --> 00:09:49,722
Michael Hunger at Neo4j already
published a blog post called Navigating
162
00:09:49,733 --> 00:09:51,697
a Neo4j Knowledge Graph with Jev.
163
00:09:52,141 --> 00:09:54,956
I'll include all the, the
links to these posts as well.
164
00:09:55,786 --> 00:09:59,725
Now, Michael didn't have access
to Jev, the GA rendition just yet.
165
00:09:59,758 --> 00:10:03,350
This was over, I guess, the end
of last week or over the weekend.
166
00:10:03,831 --> 00:10:07,267
But he started exploring before the
release, and there are three different
167
00:10:07,274 --> 00:10:08,944
operations that are available in Jev.
168
00:10:09,197 --> 00:10:12,375
The first one is a choice,
where the model will choose the
169
00:10:12,385 --> 00:10:13,862
best from a list of options.
170
00:10:14,302 --> 00:10:17,330
Then the other one is a noul
which is a score from zero to
171
00:10:17,330 --> 00:10:19,363
one of how true is a statement.
172
00:10:19,806 --> 00:10:24,349
The third option for operations is
a score, which is numeric scoring.
173
00:10:24,523 --> 00:10:27,202
I don't think the numeric scoring
was used specifically in this
174
00:10:27,204 --> 00:10:28,633
blog post, but the other two were.
175
00:10:29,016 --> 00:10:32,669
Michael walks through a demo for
how he constructed the sample
176
00:10:32,669 --> 00:10:36,778
code, along with some pros and
cons of his initial exploration.
177
00:10:37,051 --> 00:10:40,263
I definitely hope to be integrating
Jev and trying this out in my
178
00:10:40,264 --> 00:10:42,220
own projects very, very soon.
179
00:10:44,031 --> 00:10:46,983
This week, I was completely impressed
with the Meetup groups that I visited,
180
00:10:46,998 --> 00:10:50,632
and came back with a full plate of
energy and new things to explore.
181
00:10:51,005 --> 00:10:54,583
Tech is moving so quickly and doesn't
seem to stop, which was reinforced
182
00:10:54,583 --> 00:10:55,944
with this week's buzzing release.
183
00:10:56,329 --> 00:10:59,469
From upcoming events to new
things to explore, it is sure to
184
00:10:59,469 --> 00:11:01,108
keep us busy for weeks to come.
185
00:11:01,486 --> 00:11:03,924
As always, thanks for
listening, and happy coding.
