Repository navigation
docs: the investigator's pictures are five real runs on GLM-5.3, in the new look - #52
Merged
Merged
Conversation
…he new look
The sessions, audit, waterfall, memory and system pictures were real runs
on a model the work stack can no longer reach (its gateway answers 404),
taken before the boards' new look. They are retaken from five real
investigations on glm-5.3 through Zhipu's Anthropic-dialect endpoint,
priced at Z.ai's list rates ($1.40 in, $0.26 cached, $4.40 out per
million): $0.44 for the five. The prose tied to them is rewritten from
the new runs: the bill line, the report's conclusion, 52 audit calls, a
209-second waterfall, the memory it echoes. Both READMEs show the real
investigation again, above the phones.
Found while shooting: a session row carrying every flag broke each chip
inside itself ("refused" over "1"); chips now keep their words and wrap
whole.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The investigator's five real-run pictures (sessions, audit, waterfall, memory, system) were taken on
gpt-5.6-lunabefore the boards' new look, through a gateway that now answers 404. They are retaken from five real investigations onglm-5.3, through Zhipu's Anthropic-dialect endpoint, on a fresh scratch stack.HOOKPROBE_PRICE_*set to Z.ai's list rates for GLM-5.3 ($1.40 in, $0.26 cached, $4.40 out per million). The five runs cost $0.44 together, $0.05 to $0.12 each.PaymentGateway5xxon gw-3, gw-4 and gw-3's recovery, and the top-up alert. The demo memory was loaded before firing.hookprobe/docs/cost.md): the bill line, what the report concludes and how it labels inferences, the three proposed commands waiting in Approvals, 52 audit calls (4 of them the hourly selftest, refused), a 209-second waterfall over six rounds, and the memory lines the report echoes.Found while shooting: a session row carrying every flag broke each chip inside itself ("refused" over "1", "1" over "turn"). Chips now keep their words and wrap as wholes.
Also found, and fixed separately: when two runs overlap, a memory line the service auto-applies for one is blamed on the other as "rewrote its own inputs".
bash scripts/gate.sh: STACK GATE GREEN. Not deployed; nothing runs differently.