~/talk

Who Gets to Fly First Class? Value per Token in the AI Factory

links:none on file
slides:none on file

Argues that AI code generation requires hard limits: Red Hat institutes per-engineer token budgets and a centralized AI gateway to route work to the cheapest model that clears evals, defining value per token as business value delivered divided by tokens spent. Shows a queue example where business rules plus a tighter budget reorder requests, cites a 27% cheaper / 16% faster memory+mid-model combo, notes retries and failed runs inflate cost, reports 7,000 traces with ~30% showing a recurring issue, and offers first-class tokens and the trajectory miner role.

AI-written summary · about

  1. 0:00Introduction by Gene Kim
  2. 1:45Token costs and AI factory
  3. 5:30Why AI-generated PRs dropped
  4. 9:11Defining value per token
  5. 12:40Enterprise benchmarks and security patching
  6. 17:06Prioritizing work by business value
  7. 23:09Mining agent traces for value
~/search
type to search 6,518 conferences, 143,918 speakers and 192,133 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
← →
on a talk: 5 seconds back / forward
0–9
on a talk: jump to 0%–90% of the recording
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.