January 23, 2026 • 11 mins
This week has been a whirlwind. From starting a new RAG project to getting involved in other community events, there is so much to learn and do. This week had the following highlights:
🎤 Glasgow Meetup Adventures Navigating venue challenges, DJ booth speaking setups, and live coding without a mic stand—lessons in developer advocacy resilience.
🔍 RAG Experimentation Working with Quarkus to ingest unstructured data into Neo4j. Exploring filtering strategies and data model alignment for better retrieval.
💡 Live Interaction Tracer Combining naive RAG with a graph-based interaction tracer—early progress on a promising approach.
🧠Context Graphs Deep Dive Why context graphs matter for AI: documenting the "how" and "why" behind data decisions, not just snapshots in time. Perfect for providing business logic and tacit knowledge to AI systems.
Resources- Hands-on with Context Graphs and Neo4j by William Lyon
- William Lyon's podcast episode (previous month)
- Context Graphs demo application
Lots of 2026 projects kicking off—stay tuned for updates on RAG experiments, context graph implementations, and upcoming events!
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Jennifer Reif (00:05):
You are listening to the Breaktime Tech Talks podcast, a bite-sized
tech podcast for busy developers wherewe'll briefly cover technical topics, new
snippets, and more in short time blocks.
I'm your host, Jennifer Reif, anavid developer and problem solver
with special interest in data,learning, and all things technology.
(00:26):
After an international trip lastweek, I am still playing some catchup
and diving into a ton of new things.
While there weren't major epiphanieson any one particular thing this
week, I do have several irons inthe fire, if you will, so I'll share
what I'm working on and the successesand or roadblocks that I'm hitting.
First off, I wanna give ahuge shout out to the Glasgow
(00:46):
meetup where I spoke last week.
This was a really great event, and itwas a great group of attendees with
a little bit of an unusual venue.
There were lots of venues that fellthrough before the event, so the
organizers were scrambling a bit.
I felt really sorry for themtrying to pull things together.
When the venue kept changingand falling through on them.
(01:08):
However, what they ended upwith was a really nice space.
It had a nice drinkservice, and plenty of room.
It was a very long room, but not as wide.
You could tell this would have been reallywell set up for things like professional
or personal parties and fun, fancy events.
However, the speaker accommodationswere a little bit unusual.
(01:29):
They were up on a platformand off to the side.
It felt a little bit awkward for meto stand so far above the audience,
so I had to plug in my laptop up and,and up to the right, a few stairs up.
And stand kind of behind this wall andyou could tell it was meant for a DJ
where it's supposed to hide the operator.
I'm a relatively petite person, sowith that kind of large wall, me
(01:51):
being up so high, it was a littlebit hard for me to see over the edge
of it and just for people to see me.
There also was an LED screen,which was really beautiful.
I was really excited about it, pluggedeverything in, but I realized later on
that it actually was a little bit hardto read code from it, especially for
(02:11):
anyone sitting in the back of the room.
We fiddled around with the, screendisplay, resolution and the font sizes and
things to make that a little bit easier.
A little bit of a hurdlethere, but we managed it and it
worked pretty well after that.
There also was no wifi.
So had a few little hurdles totry to find a roundabout way.
We were able to get someconnection, set up and going,
(02:34):
and I was able to do the demo.
It just took a little bit of,finagling and adjustment on my
end to get everything working.
Then there was a handheld mic.
Now I've seen lots of venues thathave this, before where there's no
mic stand or there's no lapel mic.
Not an huge ordeal, especially if you'rejust working from slides, but for someone,
who live codes quite a bit, that getsrather tricky to try to project my
(02:59):
voice to the back of the room while I'vegot both hands on the keyboard trying
to type and talk at the same time.
Thankfully, I had an amazing colleaguewho acted as my mic stand and helped
me continue, moving through thepresentation while he held the mic
so that I could get my code writtenand both hands on the keyboard.
All in all, the event was quite anadventure, but we made it in the end.
(03:22):
I'm incredibly grateful to attendees andorganizer for putting everything together.
I would definitely go backin a heart beat as well.
The organizer was fantasticand the attendees were all
attentive and interested.
It was a great event.
As advocates, though, this is a greatlesson that I had planned and have
dealt with many of these challengesbefore, but it's always entertaining
(03:43):
afterward, of course, to see what endsup cropping up during a live event.
With battling some of thesethings, I have the attitude
that I will prevail in the end.
Sometimes you just have to fightfor it a little more than others.
But that's all right.
Overall, still a great event andhope to go back in the future.
The other things I've been working onthis week is that I've been playing around
(04:03):
with Corcus and loading in unstructureddata to Neo4j, doing some ingestion
and then some retrieval on there.
I'm trying a few differentfiltering approaches, and I
haven't had a lot of success yet.
I think it's just a matter of aligningthe data structure, the data model, if
you will, to the retrieval strategy, tothe ingestion piece and making sure all
(04:26):
of those sync up well and I just haven'tfound a good fit for all of that yet.
I might need the combinationof rag and text-to-Cypher.
Right now I'm just working with PlainRag, but I am playing around with
some more of the filtering things withRAG before I dive into other options.
I'm also trying to figure outa good structure for the data
model before I even ingest it.
(04:47):
Basically how the model will lookonce I do the ingestion and then
right now it feels very disjointed.
This seems like a good case for naiveRag right now, but there are a few
things also that I would like to add,the relationships and the connections
to hopefully use later on down the road.
There are, I am also wanting tocombine this naive RAG with a
(05:12):
live interaction tracer, which isdefinitely a great graph use case.
That's why I've left allof the data in the graph.
For right now, I'm still finetuning how this Live Interaction
Tracer is going to work, but we arestarting to make some progress there.
I hope to have some more tips and thoughtson this approach in the near future.
There's also been a lot of randommeetings with folks just wanting
(05:34):
to kick off projects or events.
Is anyone else feeling this as well?
With the start of the year?
We are now late enough past January 1stthat everyone's back, but not late enough
where people are quite heads down yet.
So there's a lot of planning phaseand strategizing and kicking.
Off projects.
So I've just ended up with a lot ofmeetings, dropped into my schedule to
(05:56):
talk about future events or projectsor things that we want done this year.
There should be some really fun andinteresting things though coming out of
these, so I will keep you posted on that.
For the content piece for this week,I read a blog from Will Lyon, which if
you've been around this podcast, youknow that, he actually was my guest
(06:16):
in last month's podcast or one ofthe, one of the podcasts last month.
And he recently came out with ablog talking about context graphs.
And if you've been around the AIspace or are keeping up with the
latest and greatest, this is a topicthat has come up relatively recently.
And it's starting to makea splash in the industry.
This idea of context graphs and WillLyon wrote a blog post about how,
(06:42):
context graphs and Neo4j work together.
It's called Hands-on withContext Graphs and Neo4j.
First off, the blog startswalking through what is a
context graph and why it matters.
I'll give a short little definitionhere and then, continue on.
But traditional databases have donereally well at capturing a snapshot of
(07:02):
data in time, but they don't really tellyou how and why that data got there.
Context graphs can help documentthe how and why of those decisions
so that you can understand why thedata looks the way that it does.
This can be really powerful for AIespecially, because AI doesn't understand
(07:23):
and can't really read between the linesto get the underlying context behind
those decisions, context graphs canhelp provide that tacit knowledge.
The implicit human held informationin our brains or business rules or
whatever it might happen to be, andadd that information to the data.
(07:45):
The blog post steps through Anexample of a credit line increase
for a customer, whether you shouldapprove it or not approve it.
And the example here is that it cango through and do a customer search,
calculate risk score, look at theemployer and the risks there, any recent
transactions, and figure out the decision.
(08:07):
And then record the basis of thatdecision, Hey, we are rejecting
this customer's credit line increasebecause of these particular reasons
and documenting all of that.
This is something that previously youwould've built a, an application or
a system, and then an SME would'vebeen required to come in and read the
(08:28):
code and analyze and then convey, hey,this is why we made these decisions.
Or there would've been maybedocumentation or notes to say, here
are the business rules that we usedfor this calculation or for this logic.
But it's never really beenpossible or considered until now
to start documenting these things.
(08:49):
But because now we have an AI, we needsome way to provide those business
decisions and that logic and that SMEknowledge to an external source that's not
human to human communication, and this iswhere something like a context graph can
help provide and, and fill these gaps.
It's a way that you can document thedecisions, the processes, the logic rules
(09:14):
that go into why did we choose this, andhow did this customer end up with a credit
line rejection at this point in time?
You can also document things overa long course of time, right?
Because you can look at decisionsover time and say, well.
Now this customer has actuallyimproved their credit score.
They've been a longstanding good citizen,et cetera, and now we can actually
(09:37):
approve that credit line increasedown the road, even though we couldn't
back then because of X, Y, Z reasons.
Now those reasons are no longer valid.
Now we can approve thiscredit line increase.
So even just things that change overtime, you can document all of this in
a graph and follow that path and thatjourney for understanding when and
(09:57):
where and why a customer might haveaccess to some things and not to others.
There's also lots of other use case casesfor context graphs, but this was just
one that I think is really relevant andmakes sense that previously you wouldn't
have access or you wouldn't be able tounderstand the decisions that went in
(10:18):
behind that logic without an SME present.
And now we're actually documenting this inthe database itself using context graphs.
The blog post walks through asolution architecture, some of
the technical details, includingthings like the algorithms that were
run and the data model and so on.
And then also provides a demo with linksthat anyone can go out and replicate the
(10:40):
demo application and play around with it.
Or , just go directly to the hostedrendition of the demo app, and look at
and understand how context graphs work.
This was a busy week, but I'msuper excited to see where this
month and 2026 is taking me.
Just with all of the projectsand things that are spinning up.
(11:01):
There's a lot of excitement and a lot ofinterest and lots of things to work on.
I will keep you posted.
As always, thanks forlistening and happy coding.
You are listening to the Breaktime Tech Talks podcast, a bite-sized
tech podcast for busy developers wherewe'll briefly cover technical topics, new
snippets, and more in short time blocks.
I'm your host, Jennifer Reif, anavid developer and problem solver
with special interest in data,learning, and all things technology.
(00:26):
After an international trip lastweek, I am still playing some catchup
and diving into a ton of new things.
While there weren't major epiphanieson any one particular thing this
week, I do have several irons inthe fire, if you will, so I'll share
what I'm working on and the successesand or roadblocks that I'm hitting.
First off, I wanna give ahuge shout out to the Glasgow
(00:46):
meetup where I spoke last week.
This was a really great event, and itwas a great group of attendees with
a little bit of an unusual venue.
There were lots of venues that fellthrough before the event, so the
organizers were scrambling a bit.
I felt really sorry for themtrying to pull things together.
When the venue kept changingand falling through on them.
(01:08):
However, what they ended upwith was a really nice space.
It had a nice drinkservice, and plenty of room.
It was a very long room, but not as wide.
You could tell this would have been reallywell set up for things like professional
or personal parties and fun, fancy events.
However, the speaker accommodationswere a little bit unusual.
(01:29):
They were up on a platformand off to the side.
It felt a little bit awkward for meto stand so far above the audience,
so I had to plug in my laptop up and,and up to the right, a few stairs up.
And stand kind of behind this wall andyou could tell it was meant for a DJ
where it's supposed to hide the operator.
I'm a relatively petite person, sowith that kind of large wall, me
(01:51):
being up so high, it was a littlebit hard for me to see over the edge
of it and just for people to see me.
There also was an LED screen,which was really beautiful.
I was really excited about it, pluggedeverything in, but I realized later on
that it actually was a little bit hardto read code from it, especially for
(02:11):
anyone sitting in the back of the room.
We fiddled around with the, screendisplay, resolution and the font sizes and
things to make that a little bit easier.
A little bit of a hurdlethere, but we managed it and it
worked pretty well after that.
There also was no wifi.
So had a few little hurdles totry to find a roundabout way.
We were able to get someconnection, set up and going,
(02:34):
and I was able to do the demo.
It just took a little bit of,finagling and adjustment on my
end to get everything working.
Then there was a handheld mic.
Now I've seen lots of venues thathave this, before where there's no
mic stand or there's no lapel mic.
Not an huge ordeal, especially if you'rejust working from slides, but for someone,
who live codes quite a bit, that getsrather tricky to try to project my
(02:59):
voice to the back of the room while I'vegot both hands on the keyboard trying
to type and talk at the same time.
Thankfully, I had an amazing colleaguewho acted as my mic stand and helped
me continue, moving through thepresentation while he held the mic
so that I could get my code writtenand both hands on the keyboard.
All in all, the event was quite anadventure, but we made it in the end.
(03:22):
I'm incredibly grateful to attendees andorganizer for putting everything together.
I would definitely go backin a heart beat as well.
The organizer was fantasticand the attendees were all
attentive and interested.
It was a great event.
As advocates, though, this is a greatlesson that I had planned and have
dealt with many of these challengesbefore, but it's always entertaining
(03:43):
afterward, of course, to see what endsup cropping up during a live event.
With battling some of thesethings, I have the attitude
that I will prevail in the end.
Sometimes you just have to fightfor it a little more than others.
But that's all right.
Overall, still a great event andhope to go back in the future.
The other things I've been working onthis week is that I've been playing around
(04:03):
with Corcus and loading in unstructureddata to Neo4j, doing some ingestion
and then some retrieval on there.
I'm trying a few differentfiltering approaches, and I
haven't had a lot of success yet.
I think it's just a matter of aligningthe data structure, the data model, if
you will, to the retrieval strategy, tothe ingestion piece and making sure all
(04:26):
of those sync up well and I just haven'tfound a good fit for all of that yet.
I might need the combinationof rag and text-to-Cypher.
Right now I'm just working with PlainRag, but I am playing around with
some more of the filtering things withRAG before I dive into other options.
I'm also trying to figure outa good structure for the data
model before I even ingest it.
(04:47):
Basically how the model will lookonce I do the ingestion and then
right now it feels very disjointed.
This seems like a good case for naiveRag right now, but there are a few
things also that I would like to add,the relationships and the connections
to hopefully use later on down the road.
There are, I am also wanting tocombine this naive RAG with a
(05:12):
live interaction tracer, which isdefinitely a great graph use case.
That's why I've left allof the data in the graph.
For right now, I'm still finetuning how this Live Interaction
Tracer is going to work, but we arestarting to make some progress there.
I hope to have some more tips and thoughtson this approach in the near future.
There's also been a lot of randommeetings with folks just wanting
(05:34):
to kick off projects or events.
Is anyone else feeling this as well?
With the start of the year?
We are now late enough past January 1stthat everyone's back, but not late enough
where people are quite heads down yet.
So there's a lot of planning phaseand strategizing and kicking.
Off projects.
So I've just ended up with a lot ofmeetings, dropped into my schedule to
(05:56):
talk about future events or projectsor things that we want done this year.
There should be some really fun andinteresting things though coming out of
these, so I will keep you posted on that.
For the content piece for this week,I read a blog from Will Lyon, which if
you've been around this podcast, youknow that, he actually was my guest
(06:16):
in last month's podcast or one ofthe, one of the podcasts last month.
And he recently came out with ablog talking about context graphs.
And if you've been around the AIspace or are keeping up with the
latest and greatest, this is a topicthat has come up relatively recently.
And it's starting to makea splash in the industry.
This idea of context graphs and WillLyon wrote a blog post about how,
(06:42):
context graphs and Neo4j work together.
It's called Hands-on withContext Graphs and Neo4j.
First off, the blog startswalking through what is a
context graph and why it matters.
I'll give a short little definitionhere and then, continue on.
But traditional databases have donereally well at capturing a snapshot of
(07:02):
data in time, but they don't really tellyou how and why that data got there.
Context graphs can help documentthe how and why of those decisions
so that you can understand why thedata looks the way that it does.
This can be really powerful for AIespecially, because AI doesn't understand
(07:23):
and can't really read between the linesto get the underlying context behind
those decisions, context graphs canhelp provide that tacit knowledge.
The implicit human held informationin our brains or business rules or
whatever it might happen to be, andadd that information to the data.
(07:45):
The blog post steps through Anexample of a credit line increase
for a customer, whether you shouldapprove it or not approve it.
And the example here is that it cango through and do a customer search,
calculate risk score, look at theemployer and the risks there, any recent
transactions, and figure out the decision.
(08:07):
And then record the basis of thatdecision, Hey, we are rejecting
this customer's credit line increasebecause of these particular reasons
and documenting all of that.
This is something that previously youwould've built a, an application or
a system, and then an SME would'vebeen required to come in and read the
(08:28):
code and analyze and then convey, hey,this is why we made these decisions.
Or there would've been maybedocumentation or notes to say, here
are the business rules that we usedfor this calculation or for this logic.
But it's never really beenpossible or considered until now
to start documenting these things.
(08:49):
But because now we have an AI, we needsome way to provide those business
decisions and that logic and that SMEknowledge to an external source that's not
human to human communication, and this iswhere something like a context graph can
help provide and, and fill these gaps.
It's a way that you can document thedecisions, the processes, the logic rules
(09:14):
that go into why did we choose this, andhow did this customer end up with a credit
line rejection at this point in time?
You can also document things overa long course of time, right?
Because you can look at decisionsover time and say, well.
Now this customer has actuallyimproved their credit score.
They've been a longstanding good citizen,et cetera, and now we can actually
(09:37):
approve that credit line increasedown the road, even though we couldn't
back then because of X, Y, Z reasons.
Now those reasons are no longer valid.
Now we can approve thiscredit line increase.
So even just things that change overtime, you can document all of this in
a graph and follow that path and thatjourney for understanding when and
(09:57):
where and why a customer might haveaccess to some things and not to others.
There's also lots of other use case casesfor context graphs, but this was just
one that I think is really relevant andmakes sense that previously you wouldn't
have access or you wouldn't be able tounderstand the decisions that went in
(10:18):
behind that logic without an SME present.
And now we're actually documenting this inthe database itself using context graphs.
The blog post walks through asolution architecture, some of
the technical details, includingthings like the algorithms that were
run and the data model and so on.
And then also provides a demo with linksthat anyone can go out and replicate the
(10:40):
demo application and play around with it.
Or , just go directly to the hostedrendition of the demo app, and look at
and understand how context graphs work.
This was a busy week, but I'msuper excited to see where this
month and 2026 is taking me.
Just with all of the projectsand things that are spinning up.
(11:01):
There's a lot of excitement and a lot ofinterest and lots of things to work on.
I will keep you posted.
As always, thanks forlistening and happy coding.
Popular Podcasts
NFL Daily with Gregg Rosenthal
Gregg Rosenthal and a rotating crew of elite NFL Media co-hosts, including Patrick Claybon, Colleen Wolfe, Steve Wyche, Nick Shook and Jourdan Rodrigue of The Athletic get you caught up daily on all the NFL news and analysis you need to be smarter and funnier than your friends.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.