BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//programme.europython.eu//europython-2026//talk//MJTZ7A
BEGIN:VTIMEZONE
TZID:CET
BEGIN:STANDARD
DTSTART:20001029T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-europython-2026-MJTZ7A@programme.europython.eu
DTSTART;TZID=CET:20260716T124500
DTEND;TZID=CET:20260716T131500
DESCRIPTION:Artificial intelligence has become deeply embedded in modern Py
 thon development workflows. From generating backend services to refactorin
 g production systems\, large language models now influence how software is
  designed\, implemented\, and shipped. While these tools offer significant
  productivity gains\, many teams are discovering a less visible cost: code
  is increasingly produced faster than architectural decisions can be revie
 wed\, validated\, and governed.\n\nIn real production environments\, parti
 cularly those built on distributed services\, data pipelines\, and cloud i
 nfrastructure\, unstructured AI-driven development often leads to predicta
 ble outcomes. Teams encounter rising defect rates\, fragile integrations\,
  unclear system ownership\, security regressions\, and access control mist
 akes introduced by AI-generated code. Over time\, these issues surface as 
 service outages\, compliance risks\, lost clients\, and escalating mainten
 ance costs. The problem is not AI itself\, but the absence of engineering 
 structure around how it is used.\n\nThis talk examines why unsupervised 
 “vibe coding” fails at scale and how development teams can adopt a dis
 ciplined\, AI-assisted development model that improves both velocity and r
 eliability. Drawing from real-world backend systems\, I will present pract
 ical techniques for embedding AI across the Software Development Life Cycl
 e — including structured design inputs\, architecture validation\, autom
 ated reviews\, testing strategies\, and continuous quality controls.\n\nTo
  ground the discussion in reality\, the session includes a concrete produc
 tion case study from a rapidly developed\, AI-assisted Python system. Star
 ting from access to a single project\, I was able to traverse service boun
 daries and gain visibility into multiple cloud environments and internal r
 epositories across both AWS and GCP. The root cause was not a single vulne
 rability\, but a chain of small\, AI-generated decisions: overly broad per
 missions\, copied infrastructure patterns\, missing ownership boundaries\,
  and unreviewed assumptions propagated across services. The result was a s
 ystem that appeared to move quickly\, but ultimately required emergency re
 mediation\, delayed releases\, and loss of trust.\n\nThe talk concludes by
  addressing a common misconception: a 50% increase in coding speed does no
 t translate into 50% faster product delivery. Without governance\, the opp
 osite is often true.\n\nThe session is aimed at Python developers\, techni
 cal leads\, and architects responsible for production systems. Familiarity
  with Python backend development is recommended\, but no prior experience 
 with AI tooling is required.
DTSTAMP:20260524T130833Z
LOCATION:Conference Hall Complex (S4)
SUMMARY:The hidden cost of vibe coding - Sebastian Burzyński
URL:https://programme.europython.eu/europython-2026/talk/MJTZ7A/
END:VEVENT
END:VCALENDAR