BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.devconf.info//devconf-us-2026//talk//HGAM8H
BEGIN:VTIMEZONE
TZID:EST
BEGIN:STANDARD
DTSTART:20001029T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10;UNTIL=20061029T070000Z
TZNAME:EST
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
END:STANDARD
BEGIN:STANDARD
DTSTART:20071104T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11
TZNAME:EST
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000402T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=4;UNTIL=20060402T080000Z
TZNAME:EDT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:20070311T030000
RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3
TZNAME:EDT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-devconf-us-2026-HGAM8H@pretalx.devconf.info
DTSTART;TZID=EST:20260924T134000
DTEND;TZID=EST:20260924T141500
DESCRIPTION:Most conversations about agentic AI start with the wrong questi
 on: which model should we use? Models commoditize on a six-month cycle. Th
 e team that picked the right model last quarter is one release away from b
 eing wrong\, and the next release after that. What actually differentiates
  a production agentic system isn't the model\, it's the operational layer 
 around it: tool routing\, conversation state\, cost controls\, guardrails\
 , observability\, audit trails. That layer is where the engineering compou
 nds.\n\nThis talk is for engineers building agentic systems on real backen
 ds\, not framework demos\, not greenfield AI startups\, but the boring mid
 dle where most production work lives: a service\, an ORM\, a queue\, a mod
 el API. We'll separate durable patterns that survive a model swap from dis
 posable ones you'll delete in six months as models improve\, and look at h
 ow to design the boundary between them so your stack absorbs capability ga
 ins rather than breaking.\n\nThe talk has three parts. First\, the wrong q
 uestion\, "which model?" dominates the conversation\, and what the right q
 uestion looks like. Second\, the durable layer-six patterns shown in frame
 work-free Python\, each in roughly 30 lines: tool routing\, the agentic lo
 op with bounded rounds\, conversation compaction that doesn't corrupt tool
 -call structure\, race-free cost tracking\, audit-driven observability\, a
 nd content moderation as fail-open infrastructure. Third\, the disposable 
 layer\, retry logic\, certain guardrails\, aggressive compaction\; code th
 at's load-bearing today and will be dead weight in two model generations. 
 We'll cover how to detect that a layer has aged out\, and how to keep the 
 boundary clean so removal is a deletion\, not a rewrite.\n\nWhat you'll le
 ave with: a working vocabulary for what's durable vs. disposable in your o
 wn stack (more useful\, in practice\, than the SDK / framework/harness dis
 tinction)\; a public gist of framework-free reference patterns to take hom
 e\; and a diagnostic for spotting the parts of your current system that ar
 e about to age out.\n\nWhat this talk deliberately avoids: framework recom
 mendations or comparisons (this talk is about the layer frameworks abstrac
 ted over)\, live model demos (capability differences across frontier model
 s are not the point)\, and product case studies. Patterns are presented st
 ack-agnostically\; production grounding is implicit\, not the spine.\n\nAu
 dience: backend and platform engineers shipping agentic features in produc
 tion\; technical leads making architecture decisions about whether to adop
 t agent frameworks. Prerequisites: comfort reading Python and basic famili
 arity with calling an LLM API. No prior agent framework experience require
 d\; this talk works equally well for engineers who've never used LangChain
  and for engineers ripping it out.
DTSTAMP:20260727T174744Z
LOCATION:Ladd Room (Capacity 170)
SUMMARY:The Model Is a Commodity. The Harness Is the Product - Shubhank Sax
 ena
URL:https://pretalx.devconf.info/devconf-us-2026/talk/HGAM8H/
END:VEVENT
END:VCALENDAR