The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →The QCon AI New York 2026 session titled “Production Patterns for Agents, Evals, and LLM Costs” is not confirmed by an official event page or matching session listing. Its speaker, date, schedule, and abstract remain unverified. Related conference material offers useful context about agent evaluation and operations, but it does not establish what this QCon session will cover.
What is known about the QCon session
No official QCon AI New York 2026 event page or matching session listing is available to substantiate the title. That means there is no verified speaker, session time, event date, abstract, or confirmation that the talk is on the program. The title alone is not enough to establish those details or to treat the session as announced.
As an Amazon Associate I earn from qualifying purchases.
Readers looking for attendance or speaker information should rely on a QCon event schedule or session page once one is available. Until then, descriptions of the talk’s contents would be speculation. In particular, no quotations or specific recommendations can be attributed to a QCon speaker.
What related conference material says about production agents
Separate Linux Foundation program entries describe recurring engineering concerns for agent systems. They are useful background for understanding the title’s themes, but they concern other sessions—not this QCon listing.
Evaluate more than the final answer
One program entry describes measuring agent behavior at several levels: the final outcome, the trajectory taken to reach it, tool accuracy, and intermediate sub-goals. It also distinguishes deterministic scorers from explainable LLM judges and mentions production A/B tests. These are evaluation dimensions identified in that separate program, not a confirmed framework or recommendation from QCon. See the Linux Foundation program entry on evaluation.
Use traces to investigate workflow changes
Another session summary discusses tracing and evaluation of agent workflows, including context use and output formats as models and harnesses change. That points to a practical observability question: can a team inspect enough of an agent’s work to understand why a result changed? The summary does not provide a QCon-specific method or establish a particular cost-reduction technique. Read the separate Linux Foundation entry on tracing and evaluation.
Account for the work of self-hosting
A third session summary names routing, retries, context management, quantization tradeoffs, observability, and failure recovery as operational issues for self-hosted agents. These concerns suggest that deployment choices affect more than model selection: teams also need to manage how requests are routed, failures handled, and system behavior observed. The entry does not establish that QCon’s session addresses self-hosting or compares hosted and self-hosted systems. See the Linux Foundation entry on self-hosted agent operations.
What “LLM costs” does—and does not—tell readers
“LLM costs” appears in the supplied title, but there are no verified figures, pricing details, measurements, or cost-control tactics tied to the QCon session. The adjacent material raises context usage and operational choices as topics teams may examine, but does not quantify their cost or show that any one approach saves money. Readers should not infer a benchmark, a cost comparison, or a promised saving from the title.
If a confirmed abstract becomes available, useful details to check include what cost is being measured, which workload and deployment it applies to, and whether the speaker reports measured results or proposes a general approach. Without those specifics, a cost claim cannot be meaningfully compared across systems.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to interpret coverage before the program is confirmed
- Treat the session title as unverified until it appears in an official QCon schedule or event page.
- Do not assume the Linux Foundation sessions are part of QCon or that their speakers, methods, or claims belong to the named talk.
- Look for an official abstract before drawing conclusions about the session’s scope, and for primary materials before attributing detailed technical claims or quotations.
The distinction matters: agent evaluation, observability, self-hosting, and cost management are related production concerns, but the available program entries do not show that a particular QCon speaker will discuss them. At present, the defensible account is limited to the unverified listing title and the separate conference context above.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →
