Everything Is Showbiz: Lessons from a PHP + AI Side Project

This talk grew out of a side project that got wildly out of hand. I set out to build a quick Laravel app to make one of my favourite podcasts searchable - just a weekend hack, a bit of "vibe coding" to scratch an itch. Instead, it became EverythingIsShowbiz.com, a fully fledged PHP + AI transcript engine with semantic search. What started as an experiment turned into a practical crash course in where AI actually fits into PHP workflows: from managing GPU-heavy transcription jobs, to cleaning up potentially-libellous misheard phrases, making AI output trustworthy at scale, and all on a side project budget!

Along the way, I learned a lot about the gap between AI hype and engineering reality. We'll explore the messy middle - where Laravel pipelines replace notebooks, admin tools become sanity savers, and side projects quietly teach us the same lessons as production systems. This talk is for anyone who's ever been asked, "can you do that... but add AI in?" and wants to know what that really looks like in PHP.

A small reminder that PHP still belongs in the future - even when that future comes wrapped in AI.

Transcript

1. Everything Is Showbiz: Lessons from a PHP + AI Side Project Paul Conroy / @conroyp
2. Paul Conroy From Dublin, Ireland Started playing with the web 30+ years ago (Notepad, Frontpage & Geocities!) CTO at Square1 conroyp.com / @conroyp 👴 🌍 🇮🇪
4. • Interview comedians and celebrities

5. • Interview comedians and celebrities
 • Ask them what they did yesterday

6. • Interview comedians and celebrities
 • Ask them what they did yesterday
 • That’s it!
10. What do I want?
11. • Some kind of episode transcription
 What do I want?
12. • Some kind of episode transcription
 • Searchable
 What do I want?
13. • Some kind of episode transcription
 • Searchable
 • Easily share specific parts
 What do I want?
14. • Some kind of episode transcription
 • Searchable
 • Easily share specific parts
 • That’s it! What do I want?
15. • Some kind of episode transcription
 • Searchable
 • Easily share specific parts
 • That’s it! What do I want?
16. • Some kind of episode transcription
 • Searchable
 • Easily share specific parts
 • That’s it! What do I want?
17. • Some kind of episode transcription
 • Searchable
 • Easily share specific parts
 • That’s it! What do I want?
18. INGEST TRANSCRIBE SERVE A Simple Plan!
19. Things I already know how to do Challenges in this project Audio Transcription Semantic Search Dynamic front end Model evaluation RSS Parsing Queues Cost-effective hosting Saying “we’ll tidy that up later Convincing myself it’s “MVP” 
 
 Making confusing Venn diagrams
20. I needed: I get: Something to run an ingest twice a week Scheduler Process files in the background Queues Try different AI tools Service classes & container A simple UI to browse results Blade, Alpine Deploy & run at low cost Cache, config, cheap VPS
21. TRANSCRIBE SERVE INGEST A Simple Plan
22. Let’s get started!
23. • Design the entities / models Let’s get started!
24. • Design the entities / models • Map out the Controllers, Jobs, Services Let’s get started!
25. • Design the entities / models • Map out the Controllers, Jobs, Services • Set up config and environments Let’s get started!
26. • Design the entities / models • Map out the Controllers, Jobs, Services • Set up config and environments • Set up test / validation rig Let’s get started!
27. • Design the entities / models • Map out the Controllers, Jobs, Services • Set up config and environments • Set up test / validation rig • Fix problems with test / validation rig Let’s get started!
28. • Design the entities / models • Map out the Controllers, Jobs, Services • Set up config and environments • Set up test / validation rig • Fix problems with test / validation rig • Build queue infrastructure Let’s get started!
29. • Design the entities / models • Map out the Controllers, Jobs, Services • Set up config and environments • Set up test / validation rig • Fix problems with test / validation rig • Build queue infrastructure • Fix problems with queues not running Let’s get started!
30. • Design the entities / models • Map out the Controllers, Jobs, Services • Set up config and environments • Set up test / validation rig • Fix problems with test / validation rig • Build queue infrastructure • Fix problems with queues not running • …. Let’s get started!
37. • Design the entities / models • Map out the Controllers, Jobs, Services • Set up config and environments • Set up test / validation rig • Fix problems with test / validation rig • Build queue feature • Fix problems with queues not running • …. Let’s get started!
41. TRANSCRIBE SERVE INGEST A Simple Plan
42. Whisper curl --request POST \ --url https://api.openai.com/v1/audio/transcriptions \ --header "Authorization: Bearer $OPENAI_API_KEY" \ --header 'Content-Type: multipart/form-data' \ --form file=@/path/to/file/audio.mp3 \ --form model=gpt-4o-transcribe { "text": “Imagine the wildest idea that you've ever had, and you're curious about how it might scale to something that's a 100, a 1,000 times bigger.....” } • OpenAI API • Fast! • Supports prompting • Post-processing https://developers.openai.com/api/docs/guides/speech-to-text/
43. Whisper curl --request POST \ --url https://api.openai.com/v1/audio/transcriptions \ --header "Authorization: Bearer $OPENAI_API_KEY" \ --header 'Content-Type: multipart/form-data' \ --form file=@/path/to/file/audio.mp3 \ --form model=gpt-4o-transcribe { "text": “Imagine the wildest idea that you've ever had, and you're curious about how it might scale to something that's a 100, a 1,000 times bigger.....” } • OpenAI API • Fast! • Supports prompting • Post-processing https://developers.openai.com/api/docs/guides/speech-to-text/
44. Whisper curl --request POST \ --url https://api.openai.com/v1/audio/transcriptions \ --header "Authorization: Bearer $OPENAI_API_KEY" \ --header 'Content-Type: multipart/form-data' \ --form file=@/path/to/file/audio.mp3 \ --form model=gpt-4o-transcribe { "text": “Imagine the wildest idea that you've ever had, and you're curious about how it might scale to something that's a 100, a 1,000 times bigger.....” } • OpenAI API • Fast! • Supports prompting • Post-processing https://developers.openai.com/api/docs/guides/speech-to-text/
45. https://github.com/ggml-org/whisper.cpp
46. • Runs locally! • Plain C/C++ implementation • No python required • Different models can be used • CPU/GPU-hungry https://github.com/ggml-org/whisper.cpp
47. • Large variations in model quality • --max-context keeps internal consistency high • --prompt can help with noun recognition • Prompt too long, and the quality degrades - a balancing act!
49. https://github.com/openai/whisper/discussions/194
50. • Sporadic issues with proper nouns • Challenges with punctuation & run-on sentences • Occasionally gets a little confused… https://github.com/openai/whisper/discussions/194
51. Symphony of the butt: https://everythingisshowbiz.com/?episode=323&segment=390151
52. • Transcription breaks on pause
 • Lots of short segments • Not great for search!

55. • Transcription breaks on pause
 • Lots of short segments • Not great for search!

56. • Transcription breaks on pause
 • Lots of short segments • Not great for search!

57. TRANSCRIBE SERVE INGEST A Simple Plan
58. TRANSCRIBE SERVE REFINE TEXT INGEST A Simple Plan
59. INGEST TRANSCRIBE SERVE REFINE TEXT A Simple(-ish) Plan
60. What’s in a name? • Models trained on “mid-Atlantic” accent
 • Show features strong regional accents
 • Names pronounced differently in each episode • Post-transcription processing needed
65. Actual Conversation
66. Actual Conversation Original Transcription
67. Responsibility • Content is AI-generated, but responsibility is ours
 • Need guardrails to prevent incorrect publication • Errors already costing companies
68. Responsibility • Content is AI-generated, but responsibility is ours
 • Need guardrails to prevent incorrect publication • Errors already costing companies https://www.theguardian.com/world/2024/feb/16/air-canada-chatbot-lawsuit
75. INGEST TRANSCRIBE SERVE REFINE TEXT A Simple(-ish) Plan
77. 🎉🥳
82. INGEST TRANSCRIBE SERVE REFINE TEXT A Simple(-ish) Plan
83. INGEST TRANSCRIBE SERVE REFINE TEXT SEARCH SERVE A Simple(-ish) Plan
84. INGEST TRANSCRIBE SERVE REFINE TEXT SEARCH A Plan
85. Semantic Search • Keyword search asks: “Does this text contain these words?” • Semantic search asks: “Is this about the same thing?” • It tolerates synonyms, paraphrasing, and noise • Under the hood: high-dimensional math
86. Semantic Search • Keyword search asks: “Does this text contain these words?” • Semantic search asks: “Is this about the same thing?” • It tolerates synonyms, paraphrasing, and noise • Under the hood: high-dimensional math • ✨✨ Embeddings! ✨✨
87. • Complex data reduced to simpler information [159, 837, 528, 162, …] • The larger the embedding, the more “meaning” captured
 • Similar words or phrases are close to each other in this multi-dimensional space
88. • Complex data reduced to simpler information [159, 837, 528, 162, …] • The larger the embedding, the more “meaning” captured
 • Similar words or phrases are close to each other in this multi-dimensional space
89. Embeddings • Postgres extension pgvector • Generate embedding for each text block • Store embedding alongside original text • Generate embedding for each query • Search by similarity https://github.com/pgvector/pgvector My interesting query
90. Embeddings • Postgres extension pgvector • Generate embedding for each text block • Store embedding alongside original text • Generate embedding for each query • Search by similarity https://github.com/pgvector/pgvector My interesting query [938, 324, 5, 234.. ]
91. Embeddings • Postgres extension pgvector • Generate embedding for each text block • Store embedding alongside original text • Generate embedding for each query • Search by similarity https://github.com/pgvector/pgvector My interesting query [938, 324, 5, 234.. ] [11, 903, 675, 837.. ]
92. Embeddings • Postgres extension pgvector • Generate embedding for each text block • Store embedding alongside original text • Generate embedding for each query • Search by similarity https://github.com/pgvector/pgvector My interesting query [938, 324, 5, 234.. ] [11, 903, 675, 837.. ]
94. Embeddings • Use local model (Ollama) or API (OpenAI) • Really basic model is ok! (text-embedding-3-small) • Makes remote APIs very cheap
95. How cheap exactly?
96. How cheap exactly?
98. Observers • Hook into model lifecycle events • Automate side-effects cleanly • Remove glue code from controllers • Keep business logic cohesive
103. Real-world users! • Tend to use really short queries • Hard to parse intent! • How to handle both?
104. Real-world users! • Tend to use really short queries • Hard to parse intent! • How to handle both?
105. Real-world users! • Tend to use really short queries • Hard to parse intent! • How to handle both?
108. Sub query, avoid recalculation
109. Boost score for exact match Sub query, avoid recalculation
114. • Search for text segments • Click segment to autoplay
 • Share deeplink • URLs always update!
115. 🤨 🤖 Issue Spotted
116. 🤨 🤖 🚀 Issue Spotted Magic Fix
117. 🤨 🤖 🚀 📦 Issue Spotted Magic Fix Change Integrated
118. 🤨 🤖 🚀 📦 Issue Spotted Magic Fix Change Integrated But then…
119. What did I expect?
120. What did I expect? • Discrete function classes / jobs

121. What did I expect? • Discrete function classes / jobs
 • Logical separation

122. What did I expect? • Discrete function classes / jobs
 • Logical separation
 • Testable
123. What did I expect? • Discrete function classes / jobs
 • Logical separation
 • Testable
124. What did I expect? • Discrete function classes / jobs
 • Logical separation
 • Testable • That’s it!
128. • Side by side of different code • Highlight “same but different” approach • AI solving from 10ft, I’m approaching this from 10,000ft
129. • Side by side of different code • Highlight “same but different” approach • AI solving from 10ft, I’m approaching this from 10,000ft
130. • Side by side of different code • Highlight “same but different” approach • AI solving from 10ft, I’m approaching this from 10,000ft
131. • Side by side of different code • Highlight “same but different” approach • AI solving from 10ft, I’m approaching this from 10,000ft
132. What’s happening? • Lots of near-duplicate code • Huge amount of dead code • AI is focusing on low-level context alone
 Working at 10ft while I’m at 10,000ft • Need to match the code to my mental model
133. Cognitive debt https://margaretstorey.com/blog/2026/02/09/cognitive-debt/
134. Cognitive debt https://margaretstorey.com/blog/2026/02/09/cognitive-debt/
135. Use AI to tidy up AI • Much tighter feedback loops • Personas ◦ “You are a senior PHP dev…”
 • Less vibes, more control • Closer to pair programming
136. Use AI to tidy up AI • Much tighter feedback loops • Personas ◦ “You are a senior PHP dev…”
 • Less vibes, more control • Closer to pair programming
137. SERVE A Plan INGEST TRANSCRIBE REFINE TEXT SEARCH
138. Workload Generation • Heavy GPU load • But only 2-3x a week • Admin tooling Public site • Effectively read-only • Updated 2-3x a week • Strong caching candidate
139. Workload Generation • Heavy GPU load • But only 2-3x a week • Admin tooling Public site • Effectively read-only • Updated 2-3x a week • Strong caching candidate 💸💸💸
140. Workload Generation • Heavy GPU load • But only 2-3x a week • Admin tooling Public site • Effectively read-only • Updated 2-3x a week • Strong caching candidate 💸💸💸 💸
141. Mac Mini RSS Feed (Cheap!) VPS
142. Mac Mini RSS Feed (Cheap!) VPS
143. Mac Mini RSS Feed (Cheap!) VPS
144. Mac Mini RSS Feed (Cheap!) VPS
145. 🧑💻 Mac Mini RSS Feed (Cheap!) VPS
146. 🧑💻 Mac Mini RSS Feed (Cheap!) VPS
162. A Plan INGEST SERVE TRANSCRIBE REFINE TEXT SEARCH
163. TIDY UP A Plan INGEST SERVE TRANSCRIBE REFINE TEXT SEARCH
164. INGEST TRANSCRIBE SERVE REFINE TEXT SEARCH
165. INGEST TRANSCRIBE SERVE REFINE TEXT SEARCH TIDY UP
166. Late 2025 • Step-change in AI tooling
 • Moving from the IDE to CLI • From “check this code I wrote” to “I’ll check the code you wrote”
169. AI-Assisted Engineering • Vibe coding optimises for speed • Engineering optimises for sustainability
170. https://addyosmani.com/blog/ai-assisted-engineering-idea/ AI-Assisted Engineering • Vibe coding optimises for speed • Engineering optimises for sustainability
171. https://addyosmani.com/blog/ai-assisted-engineering-idea/ AI-Assisted Engineering • Vibe coding optimises for speed • Engineering optimises for sustainability
172. Parallel & Structured • Sub-agents with defined roles • Reusable skills
 (Laravel clean code, frontend design…) • Worktrees for isolation • Inter-agent communication 🤖 🤖 🤖 🤖 The Boss Design Security Architect
173. From prompting to planning • Interview with the model first • Capture decisions in plan.md • Make constraints explicit • Break work into checkable steps
174. From prompting to planning • Interview with the model first • Capture decisions in plan.md • Make constraints explicit • Break work into checkable steps
178. Ralphing • Models cut corners on long tasks • Hook in to finished state • “Did you really finish? Hmmmm???” • Keep going relentlessly until it’s done https://github.com/anthropics/claude-code/blob/main/plugins/ralph-wiggum/README.md
179. Cleanup Complete! • It works (mostly..) • Documented • Tested • Public https://github.com/conroyp/podcast-transcription
180. Side Projects! • Abstract away the boring boilerplate • Dopamine hit of fast feedback • Keep the momentum going • Quick to prototype & validate • From noodling on phone to live app
181. https://www.kidssudoku.com https://www.tictacgoaway.com https://www.jdcaptcha.com
182. https://www.kidssudoku.com https://www.tictacgoaway.com https://www.jdcaptcha.com
183. https://www.kidssudoku.com https://www.tictacgoaway.com https://www.jdcaptcha.com
184. Where does that leave us developers? • AI is already faster than most mid-level devs • Productivity up 5-6× → headcount down? • “Good enough” beats “well-engineered.” • Why hire juniors when AI does the boilerplate?
186. Geoffrey Hinton, “Godfather of AI”
187. “[I]f you work as a radiologist you are like Wile E. Coyote in the cartoon. You’re already over the edge of the cliff, but you haven’t yet looked down. Geoffrey Hinton, “Godfather of AI”
188. “[I]f you work as a radiologist you are like Wile E. Coyote in the cartoon. You’re already over the edge of the cliff, but you haven’t yet looked down. 
 It’s just completely obvious that in five years deep learning is going to do better than radiologists.” Geoffrey Hinton, “Godfather of AI”
189. What Happened Next? • Demand went up • Salaries increased! • The nature of the work
 changed
190. What Happened Next? • Demand went up • Salaries increased! • The nature of the work
 changed https://worksinprogress.co/issue/the-algorithm-will-see-you-now/
191. What Happened Next? • Demand went up • Salaries increased! • The nature of the work
 changed https://worksinprogress.co/issue/the-algorithm-will-see-you-now/
192. Who gets to build? • Doctors can ship apps • Lawyers can prototype tools • Contractors can automate their workflows • Founders can validate ideas without devs (?)
193. We’ve been here before • Assembly → C • C → PHP • PHP → Low-/no-code tools • Manual infra → DevOps • On-prem → Cloud
194. We’ve been here before • Assembly → C • C → PHP • PHP → Low-/no-code tools • Manual infra → DevOps • On-prem → Cloud https://addyosmani.com/blog/the-efficiency-paradox/ “Every time we’ve made it easier to write software, we’ve ended up writing exponentially more of it.”
195. What changes for us? • Execution is getting cheaper • Prototypes are cheaper • Boilerplate is getting cheaper • Judgment is not
196. What changes for us? • Execution is getting cheaper • Prototypes are cheaper • Boilerplate is getting cheaper • Judgment is not https://addyosmani.com/blog/the-efficiency-paradox/ The real question is whether we’re prepared for a world where the bottleneck shifts from “can we build this?” to “should we build this?”
198. Thank you! Conroyp.com @conroyp paul@conroyp.com 🌍 🌐 📧 https://joind.in/event/php-uk-conference-2026/everything-is-showbiz-lessons-from-a-php--ai-side-projects