research
The mainframe exit,
documented.
Strategy, cost anatomy, failure patterns, vendor matchups — and the technical foundations underneath the platform. Everything we know, written down and left open.
10 pieces · ungated · no email walls · cite freely
showing 10 of 10
Near-deterministic by design
Why Palm Ark refuses to guess: the framework that makes extraction repeatable, auditable, and safe to sign.
Gazelle — the bespoke model
Up to ~95% business-rule extraction where generic LLMs land ~30% — what a purpose-built model changes, with receipts.
The Exit Doctrine
Rehost, refactor, replatform, rewrite: the 7 Rs collapsed into the three decisions that actually matter.
What leaving the mainframe actually costs
MIPS, MLC, people, parallel-run: the full cost anatomy of staying vs leaving.
The Mainframe Exit Calculator
Your run cost vs your exit cost, modeled in the open.
Why mainframe modernization projects fail
A library of documented failure patterns — and the controls that would have caught each one.
The vendor matrix
Every serious mainframe modernization option, refereed — including the matchups we’re not in.
The RFP Kit
A complete RFP template, 40 hard questions, a PoC scope, and a board deck — no email required.
Can AI convert COBOL?
What LLMs get right, where they hallucinate, and why determinism wins audits.
The ECC 2027 clock
Your S/4HANA program has a mainframe dependency nobody scoped. The deadline doesn’t care.
nothing matches · clear a filter or two
start here
Three ways through the library.
for cios & ctos
- 1 · The Exit Doctrine
- 2 · What an exit costs
- 3 · Run your numbers
- 4 · Take the RFP Kit to procurement
forward any of these · no gate, no wall
for enterprise architects
- 1 · Near-deterministic by design
- 2 · Gazelle — the bespoke model
- 3 · The method
- 4 · PalmBench, in full
then join the room · palm neh
new to mainframes
- 1 · Can AI convert COBOL?
- 2 · Why projects fail
- 3 · The ECC 2027 clock
- 4 · Watch The Mainframe Empire
the field needs you · welcome in
quotable
Numbers you can take to a board.
Copy with the citation attached. Quote us; hold us to it.
~95%
business-rule extraction accuracy on PalmBench — vs ~30% from generic LLMs on the same program.
33–40%
of modernization scope removed before work begins, by Palm Key’s dead-code report.
#2
worldwide in product strength — second only to IBM. MarketsandMarkets, 2026.
Reading is free. So is asking.
When the reading turns into a decision, we scope a paid proof of concept — your code, your perimeter, receipts on the table.