Slide 1
Slide 1 text
6 ½ ridiculous
things to do
with quarkus
Holly Cummins
IBM + Red Hat
Slide 2
Slide 2 text
No content
Slide 3
Slide 3 text
quarkus is used in
production by:
Slide 4
Slide 4 text
quarkus is used in
production by:
• telcos
Slide 5
Slide 5 text
quarkus is used in
production by:
• telcos
• retailers
Slide 6
Slide 6 text
quarkus is used in
production by:
• telcos
• retailers
• banks
Slide 7
Slide 7 text
quarkus is used in
production by:
• telcos
• retailers
• banks
BORING!
Slide 8
Slide 8 text
No content
Slide 9
Slide 9 text
https://github.com/holly-cummins/gen-alpha-iser
Slide 10
Slide 10 text
6 ½ skibidi ohio
things to do
with quarkus
Holly Cummins
IBM + Red Hat
Slide 11
Slide 11 text
was this an easy demo?
Slide 12
Slide 12 text
was this an easy demo?
super easy
Slide 13
Slide 13 text
tell your
boss you
learned… @ServerResponseFilter
Slide 14
Slide 14 text
No content
Slide 15
Slide 15 text
https://github.com/phillip-kruger/quarkus-jokes
Slide 16
Slide 16 text
was this an easy demo?
yes, phillip krüger
wrote all of it :)
Slide 17
Slide 17 text
was this an easy demo?
yes, phillip krüger
wrote all of it :)
Slide 18
Slide 18 text
tell your
boss you
learned…
the dev ui now has
assistants to help you
write + debug code
Slide 19
Slide 19 text
No content
Slide 20
Slide 20 text
No content
Slide 21
Slide 21 text
No content
Slide 22
Slide 22 text
No content
Slide 23
Slide 23 text
No content
Slide 24
Slide 24 text
No content
Slide 25
Slide 25 text
No content
Slide 26
Slide 26 text
No content
Slide 27
Slide 27 text
https://github.com/quarkiverse/quarkus-bon-jova-rockstar
Slide 28
Slide 28 text
https://github.com/quarkiverse/quarkus-bon-jova-rockstar
Slide 29
Slide 29 text
was this an easy demo?
Slide 30
Slide 30 text
was this an easy demo?
Slide 31
Slide 31 text
was this an easy demo?
-friends don’t let friends
write compilers
Slide 32
Slide 32 text
was this an easy demo?
-friends don’t let friends
write compilers
-bytecode generation :(
Slide 33
Slide 33 text
was this an easy demo?
-friends don’t let friends
write compilers
-bytecode generation :(
-parsing :( :(
Slide 34
Slide 34 text
No content
Slide 35
Slide 35 text
No content
Slide 36
Slide 36 text
spec
Slide 37
Slide 37 text
spec
cost: 5 beers
Slide 38
Slide 38 text
spec
cost: 5 beers
time: 1 night
Slide 39
Slide 39 text
spec
cost: 5 beers
time: 1 night
implementation
Slide 40
Slide 40 text
spec
cost: 5 beers
time: 1 night
implementation
cost: $0
Slide 41
Slide 41 text
spec
cost: 5 beers
time: 1 night
implementation
cost: $0
time: 3 weeks
Slide 42
Slide 42 text
spec
spec
cost: 5 beers
time: 1 night
implementation
cost: $0
time: 3 weeks
Slide 43
Slide 43 text
spec
none, obvs
spec
cost: 5 beers
time: 1 night
implementation
cost: $0
time: 3 weeks
Slide 44
Slide 44 text
spec
none, obvs
spec
cost: 5 beers
time: 1 night
implementation
cost: $0
time: 3 weeks
Slide 45
Slide 45 text
spec
none, obvs
implementation
spec
cost: 5 beers
time: 1 night
implementation
cost: $0
time: 3 weeks
Slide 46
Slide 46 text
spec
none, obvs
implementation
cost: $36,312.03
spec
cost: 5 beers
time: 1 night
implementation
cost: $0
time: 3 weeks
Slide 47
Slide 47 text
spec
none, obvs
implementation
cost: $36,312.03
time: 3 months
spec
cost: 5 beers
time: 1 night
implementation
cost: $0
time: 3 weeks
Slide 48
Slide 48 text
gen ai: what
could possibly
go wrong?
Slide 49
Slide 49 text
No content
Slide 50
Slide 50 text
https://www.computerworld.com/article/4059383/openai-admits-ai-hallucinations-are-mathematically-inevitable-not-just-engineering-flaws.html
rate of hallucination when summarizing
public information:
Slide 51
Slide 51 text
o1: 16%
https://www.computerworld.com/article/4059383/openai-admits-ai-hallucinations-are-mathematically-inevitable-not-just-engineering-flaws.html
rate of hallucination when summarizing
public information:
Slide 52
Slide 52 text
o1: 16%
o3: 33%
https://www.computerworld.com/article/4059383/openai-admits-ai-hallucinations-are-mathematically-inevitable-not-just-engineering-flaws.html
rate of hallucination when summarizing
public information:
Slide 53
Slide 53 text
o1: 16%
o3: 33%
o4-mini: 48%
https://www.computerworld.com/article/4059383/openai-admits-ai-hallucinations-are-mathematically-inevitable-not-just-engineering-flaws.html
rate of hallucination when summarizing
public information:
Slide 54
Slide 54 text
solution: guardrails
Slide 55
Slide 55 text
No content
Slide 56
Slide 56 text
was this an easy demo?
Slide 57
Slide 57 text
was this an easy demo?
Slide 58
Slide 58 text
was this an easy demo? -adding guardrails is
simple
Slide 59
Slide 59 text
was this an easy demo? -adding guardrails is
simple
-making models do what
you want is hard
Slide 60
Slide 60 text
was this an easy demo? -adding guardrails is
simple
-making models do what
you want is hard
-hallucinations can be all
or nothing
Slide 61
Slide 61 text
gemma3 - 1 b parameters
how many bs in blueberry?
Slide 62
Slide 62 text
"Okay, let’s tackle this! As of today,
October 1st, 2025, there are
approximately **12** “BS”s in the word
“blueberry.”
It’s a surprisingly common phenomenon! 😊
Do you want to explore why this happens?"
gemma3 - 1 b parameters
how many bs in blueberry?
Slide 63
Slide 63 text
llama3.2- 1 b parameters
how many bs in blueberry?
Slide 64
Slide 64 text
"There are no \"bs\" in
Blueberry. It's spelled
B-E-L-U-E-B-R-I-E-R-Y."
llama3.2- 1 b parameters
how many bs in blueberry?
Slide 65
Slide 65 text
deepseekr1- 1.5 b parameters
when did brexit happen?
Slide 66
Slide 66 text
Hmm, the user is asking about something related to
20-30-0-3 but brexit, probably typo or20-10- something,
correction. User meant brexit or2-1- meaning Brexit. Let me
confirm that. User is20- user2 is inquiring about Brexit.
UK leaving EU thing. But what time period20 2020-3 might be
typo for20- 20 twenty021203. 20-20 or something.
Perhaps20-202201. UK left. UK left EU. UK left20 on01
Jan201202022-021201021 I think22. Wait no1. Let me212012
January22202021. So January. Yes2 user might be21
probably20 asking about that. User22. Better2 just confirm.
Brexit201 20 answer. Say20 simply22 date2. Not much else0
else0. Keep2 simple1 short1 friendly2. Okay2 polite2 user.
It seems you're asking about the date the Brexit,
right? The UK left EU on 29th January21st of
February2020221
deepseekr1- 1.5 b parameters
when did brexit happen?
Slide 67
Slide 67 text
the value of a second opinion?
chatgpt.com
Slide 68
Slide 68 text
the value of a second opinion?
chatgpt.com
the statement we’re validating
Slide 69
Slide 69 text
the value of a second opinion?
chatgpt.com
the statement we’re validating
“oh yes, definitely true”
Slide 70
Slide 70 text
the value of a second opinion?
chatgpt.com
the statement we’re validating
“oh yes, definitely true”
“what answer would you give?”
Slide 71
Slide 71 text
the value of a second opinion?
chatgpt.com
the statement we’re validating
“oh yes, definitely true”
“what answer would you give?”
but if you knew … ?
Slide 72
Slide 72 text
tell your
boss you
learned…
•model choice makes a big
difference
•“more LLM” may not be the
right solution to “LLM”
•simple guardrails not great
at detecting hallucinations
Slide 73
Slide 73 text
No content
Slide 74
Slide 74 text
No content
Slide 75
Slide 75 text
thanks to markus
eisele for the
graphics code
https://github.com/holly-cummins/meme-generator
Slide 76
Slide 76 text
was this an easy demo?
Slide 77
Slide 77 text
was this an easy demo?
Slide 78
Slide 78 text
was this an easy demo?
-writing the cli was easy
-native compilation
was easy
-native compilation +
awt + mac is very hard
Slide 79
Slide 79 text
No content
Slide 80
Slide 80 text
i tried:
Slide 81
Slide 81 text
i tried:
native awt on mac
Slide 82
Slide 82 text
i tried:
native awt on mac
liberica vm mac + tuning of
build-time and runtime classes
Slide 83
Slide 83 text
i tried:
native awt on mac
liberica vm mac + tuning of
build-time and runtime classes
graalvm community container
Slide 84
Slide 84 text
i tried:
native awt on mac
liberica vm mac + tuning of
build-time and runtime classes
graalvm community container
mandrel container
Slide 85
Slide 85 text
tell your
boss you
learned…
•picocli + quarkus is fire
•native is cool
•…but native is hard
•only go native if it adds
value
Slide 86
Slide 86 text
observability
Slide 87
Slide 87 text
No content
Slide 88
Slide 88 text
No content
Slide 89
Slide 89 text
No content
Slide 90
Slide 90 text
was this an easy demo?
-extension was easy, minecraft mod
was hard
-minecraft APIs
-lack documentation
-change between minor versions
-client-server sync of injected
behaviour is hard
Slide 91
Slide 91 text
was this an easy demo?
-extension was easy, minecraft mod
was hard
-minecraft APIs
-lack documentation
-change between minor versions
-client-server sync of injected
behaviour is hard
Slide 92
Slide 92 text
tell your
boss you
learned…
-the quarkiverse is still
very welcoming
-you can do a lot with
interceptors
-dev services are cool
Slide 93
Slide 93 text
tell your
boss you
learned…
-gen ai should integrate
into your domain model
-small models struggle
with structured data
-gen ai is not fast
Slide 94
Slide 94 text
No content
Slide 95
Slide 95 text
No content
Slide 96
Slide 96 text
was this an easy demo? -extension was easy,
minecraft mod was hard
-minecraft APIs
-lack documentation
-change between minor
versions
Slide 97
Slide 97 text
was this an easy demo? -extension was easy,
minecraft mod was hard
-minecraft APIs
-lack documentation
-change between minor
versions
Slide 98
Slide 98 text
tell your
boss you
learned…
-quarkus uses around half
the compute resources
of spring
Slide 99
Slide 99 text
tell your
boss you
learned…
-use wrk2 or hyperfoil for
load
-do not generate load
from the same machine
Slide 100
Slide 100 text
what have we
done?
Slide 101
Slide 101 text
what have we
done?
1. translating output with a filter
Slide 102
Slide 102 text
what have we
done?
1. translating output with a filter
2. transforming code with an extension
Slide 103
Slide 103 text
what have we
done?
1. translating output with a filter
2. transforming code with an extension
3. compiling rock ballads to bytecode
Slide 104
Slide 104 text
what have we
done?
1. translating output with a filter
2. transforming code with an extension
3. compiling rock ballads to bytecode
4. amplifying hallucinations
Slide 105
Slide 105 text
what have we
done?
1. translating output with a filter
2. transforming code with an extension
3. compiling rock ballads to bytecode
4. amplifying hallucinations
5. using minecraft as an observability
client
Slide 106
Slide 106 text
what have we
done?
1. translating output with a filter
2. transforming code with an extension
3. compiling rock ballads to bytecode
4. amplifying hallucinations
5. using minecraft as an observability
client
6. making memes with a cli
Slide 107
Slide 107 text
what have we
done?
1. translating output with a filter
2. transforming code with an extension
3. compiling rock ballads to bytecode
4. amplifying hallucinations
5. using minecraft as an observability
client
6. making memes with a cli
7
. confirming quarkus does a lot with
very few resources
Slide 108
Slide 108 text
gen-alpha translation services:
Asher Hodkinson
I’ll help you
but you need to
put me in the
closing credits
Slide 109
Slide 109 text
https://hollycummins.com/things-to-care-about-devoxx-be/