run 28
This commit is contained in:
Aiden
@@ -5793,3 +5793,134 @@ Best current read:
|
||||
- `AF -> E8 -> FA 50 26` is especially interesting because it suggests the
|
||||
context frame can influence **which sibling family** the selector opens, not
|
||||
just whether it opens anything at all.
|
||||
|
||||
### HE25: Do `90` And `AF` Change Downstream Behavior Too?
|
||||
|
||||
Goal:
|
||||
|
||||
- Check whether `90` and `AF` only change the *first* family opened by a
|
||||
selector, or whether they also change what happens when we answer that family.
|
||||
|
||||
Method:
|
||||
|
||||
- Use a context opener (`90` or `AF`)
|
||||
- Open a known selector branch (`E8`, `E9`, or `EC`)
|
||||
- Immediately send the exact downstream family frame that branch produced
|
||||
- Compare whether the follow-up behavior differs by opener
|
||||
|
||||
This keeps the selector and downstream family fixed while changing only the
|
||||
context opener.
|
||||
|
||||
What would count as a hit:
|
||||
|
||||
- the same downstream exact echo behaving differently under `90` vs `AF`
|
||||
- one opener causing a new second-stage family while the other falls flat
|
||||
- the opener changing whether the branch drains to heartbeat or continues
|
||||
|
||||
#### HE25a: `90 -> E8 -> exact 7A 50 26`
|
||||
|
||||
```powershell
|
||||
python scripts/serial_sequence_probe.py --port COM5 --prompt --frame "00 00 90 00 80 4A" --frame "00 00 E8 40 30 C2" --frame "07 80 7A 50 26 D1" --frame "00 00 00 00 80 DA" --frame "00 00 00 00 80 DA" --repeat 2 --frame-interval 0.20 --read-after-frame 0.30 --read-after-group 0.8 --log captures/rcp-context-90-e8-exact-7a5026.txt
|
||||
```
|
||||
|
||||
#### HE25b: `AF -> E8 -> exact FA 50 26`
|
||||
|
||||
This uses the family that `AF -> E8` actually opened.
|
||||
|
||||
```powershell
|
||||
python scripts/serial_sequence_probe.py --port COM5 --prompt --frame "00 00 AF 00 80 75" --frame "00 00 E8 40 30 C2" --frame "07 80 FA 50 26 51" --frame "00 00 00 00 80 DA" --frame "00 00 00 00 80 DA" --repeat 2 --frame-interval 0.20 --read-after-frame 0.30 --read-after-group 0.8 --log captures/rcp-context-af-e8-exact-fa5026.txt
|
||||
```
|
||||
|
||||
#### HE25c: `90 -> E9 -> exact 7A 28 D3`
|
||||
|
||||
```powershell
|
||||
python scripts/serial_sequence_probe.py --port COM5 --prompt --frame "00 00 90 00 80 4A" --frame "00 00 E9 40 30 C3" --frame "07 80 7A 28 D3 5C" --frame "00 00 00 00 80 DA" --frame "00 00 00 00 80 DA" --repeat 2 --frame-interval 0.20 --read-after-frame 0.30 --read-after-group 0.8 --log captures/rcp-context-90-e9-exact-7a28d3.txt
|
||||
```
|
||||
|
||||
#### HE25d: `AF -> E9 -> exact 7A 28 D3`
|
||||
|
||||
```powershell
|
||||
python scripts/serial_sequence_probe.py --port COM5 --prompt --frame "00 00 AF 00 80 75" --frame "00 00 E9 40 30 C3" --frame "07 80 7A 28 D3 5C" --frame "00 00 00 00 80 DA" --frame "00 00 00 00 80 DA" --repeat 2 --frame-interval 0.20 --read-after-frame 0.30 --read-after-group 0.8 --log captures/rcp-context-af-e9-exact-7a28d3.txt
|
||||
```
|
||||
|
||||
#### HE25e: `90 -> EC -> exact 7B 50 26`
|
||||
|
||||
```powershell
|
||||
python scripts/serial_sequence_probe.py --port COM5 --prompt --frame "00 00 90 00 80 4A" --frame "00 00 EC 40 30 C6" --frame "07 80 7B 50 26 D0" --frame "00 00 00 00 80 DA" --frame "00 00 00 00 80 DA" --repeat 2 --frame-interval 0.20 --read-after-frame 0.30 --read-after-group 0.8 --log captures/rcp-context-90-ec-exact-7b5026.txt
|
||||
```
|
||||
|
||||
#### HE25f: `AF -> EC -> exact 7B 50 26`
|
||||
|
||||
```powershell
|
||||
python scripts/serial_sequence_probe.py --port COM5 --prompt --frame "00 00 AF 00 80 75" --frame "00 00 EC 40 30 C6" --frame "07 80 7B 50 26 D0" --frame "00 00 00 00 80 DA" --frame "00 00 00 00 80 DA" --repeat 2 --frame-interval 0.20 --read-after-frame 0.30 --read-after-group 0.8 --log captures/rcp-context-af-ec-exact-7b5026.txt
|
||||
```
|
||||
|
||||
Recommended order:
|
||||
|
||||
1. `HE25e` (`90 -> EC -> exact 7B`)
|
||||
2. `HE25f` (`AF -> EC -> exact 7B`)
|
||||
3. `HE25c` / `HE25d` (`90/AF -> E9 -> exact 7A 28 D3`)
|
||||
4. `HE25a` / `HE25b` (`90 -> E8 -> exact 7A 50 26`, `AF -> E8 -> exact FA 50 26`)
|
||||
|
||||
That order keeps the most stateful selector first again, while still checking
|
||||
whether opener choice changes the downstream echo behavior on the simpler `E8`
|
||||
and `E9` branches.
|
||||
|
||||
### 2026-05-13 Opener-Dependent Downstream Behavior Result
|
||||
|
||||
Captures:
|
||||
|
||||
- `captures/rcp-context-90-e8-exact-7a5026.txt`
|
||||
- `captures/rcp-context-af-e8-exact-fa5026.txt`
|
||||
- `captures/rcp-context-90-e9-exact-7a28d3.txt`
|
||||
- `captures/rcp-context-af-e9-exact-7a28d3.txt`
|
||||
- `captures/rcp-context-90-ec-exact-7b5026.txt`
|
||||
- `captures/rcp-context-af-ec-exact-7b5026.txt`
|
||||
|
||||
Observed group-1 outcomes:
|
||||
|
||||
| Setup | Selector result | Exact-echo follow-up result |
|
||||
| --- | --- | --- |
|
||||
| `90 -> E8 -> exact 7A 50 26` | `07 80 7A 58 26 D9` | heartbeat only after sending `07 80 7A 50 26 D1` |
|
||||
| `AF -> E8 -> exact FA 50 26` | `07 80 7A 50 26 D1` | heartbeat only after sending `07 80 FA 50 26 51` |
|
||||
| `90 -> E9 -> exact 7A 28 D3` | `07 80 7A 28 D3 5C` | heartbeat only |
|
||||
| `AF -> E9 -> exact 7A 28 D3` | `07 80 7A 28 D3 5C` | heartbeat only |
|
||||
| `90 -> EC -> exact 7B 50 26` | `07 80 FB 50 26 50` | heartbeat only after sending `07 80 7B 50 26 D0` |
|
||||
| `AF -> EC -> exact 7B 50 26` | `07 80 7B 50 26 D0` | heartbeat only |
|
||||
|
||||
Interpretation:
|
||||
|
||||
- No opener produced a stable second-stage reply to the downstream exact echo.
|
||||
- So, for now, opener choice seems to affect the **branch entry** much more than
|
||||
the **reply-to-the-branch** behavior.
|
||||
- But opener choice definitely matters at the selector result level:
|
||||
- `90 -> E8` shifted the response from the familiar `7A 50 26` family to the
|
||||
sibling `7A 58 26 D9`
|
||||
- `90 -> EC` shifted the response from `7B 50 26 D0` to the sibling
|
||||
`FB 50 26 50`
|
||||
- `AF -> EC` preserved the `7B 50 26 D0` branch
|
||||
- `E9` looks more stable across openers than `E8` and `EC`:
|
||||
both `90` and `AF` still opened `07 80 7A 28 D3 5C`
|
||||
|
||||
Most important takeaway:
|
||||
|
||||
- opener bytes are not just yes/no "permissions" for the selector surface
|
||||
- they can change **which sibling family** a selector lands in
|
||||
- but once that family appears, echoing the expected downstream frame still did
|
||||
not create a reproducible next turn
|
||||
|
||||
Best current model:
|
||||
|
||||
- context opener selects or biases a family page
|
||||
- selector byte chooses within that opened space
|
||||
- downstream family frames still look more like readable responses than clear
|
||||
host prompts
|
||||
|
||||
Best next move:
|
||||
|
||||
- when a selector result shifts under a different opener, answer the
|
||||
**actually observed** downstream frame instead of the previously known sibling
|
||||
frame
|
||||
- the strongest immediate candidates are:
|
||||
- `90 -> E8 -> exact 07 80 7A 58 26 D9`
|
||||
- `90 -> EC -> exact 07 80 FB 50 26 50`
|
||||
|
||||
@@ -294,6 +294,11 @@ Current caution:
|
||||
`90` opened `E8`, `E9`, and `EC`; `AF` opened `E9` and `EC`, and shifted
|
||||
`E8` into the sibling `07 80 FA 50 26 51` family; bare heartbeat opened `E8`
|
||||
and `E9` but not `EC`.
|
||||
- Downstream-behavior follow-ups suggest opener choice changes the selector
|
||||
result more than the reply ladder:
|
||||
`90 -> E8` yielded sibling `07 80 7A 58 26 D9`, `90 -> EC` yielded sibling
|
||||
`07 80 FB 50 26 50`, but the downstream exact echoes still collapsed to
|
||||
heartbeat.
|
||||
|
||||
## What We Know
|
||||
|
||||
@@ -334,6 +339,8 @@ Current caution:
|
||||
- There appears to be a small family of context openers rather than a single
|
||||
magic gate byte; `EC` is stricter than `E8/E9`, and heartbeat alone was not
|
||||
enough to open it.
|
||||
- At the moment, opener bytes look more like family/page selectors than like
|
||||
things that enable a stable multi-turn conversation after the first response.
|
||||
|
||||
## What We Do Not Know
|
||||
|
||||
|
||||
@@ -142,6 +142,11 @@ Observed:
|
||||
- bare heartbeat + `E9` produced `7A 28 D3`
|
||||
- bare heartbeat + `EC` did not open `7B`; it stayed in heartbeat-family
|
||||
behavior
|
||||
- later downstream-behavior tests showed opener choice can also bias which
|
||||
sibling family appears at the selector result itself:
|
||||
- `90 + E8` produced `7A 58 26`
|
||||
- `90 + EC` produced `FB 50 26`
|
||||
- `AF + EC` preserved `7B 50 26`
|
||||
|
||||
Nuance:
|
||||
|
||||
@@ -151,6 +156,8 @@ Nuance:
|
||||
earlier `A0 + E9` runs produced `07 80 7A 28 D3 5C`.
|
||||
- `EC` appears stricter than `E8/E9`, because heartbeat alone was enough for
|
||||
`E8/E9` but not for `EC`.
|
||||
- opener choice has a stronger observed effect on branch entry than on
|
||||
downstream reply handling
|
||||
|
||||
Confidence: high that a context-opener family exists; medium-high on the exact
|
||||
rules for each selector.
|
||||
@@ -198,7 +205,8 @@ What we know:
|
||||
- `A0`
|
||||
- `90`
|
||||
- bare heartbeat
|
||||
- `AF` shifted this branch to sibling `FA 50 26 51`
|
||||
- `AF` has at least once shifted this branch to sibling `FA 50 26 51`
|
||||
- `90` has at least once shifted this branch to sibling `7A 58 26 D9`
|
||||
|
||||
Read:
|
||||
|
||||
@@ -270,6 +278,7 @@ Read:
|
||||
- `7B` is the most stable downstream family on this branch
|
||||
- `FB` is real but currently looks like a sibling observation, not a clear reply
|
||||
target
|
||||
- opener choice can bias `EC` between `7B` and `FB`
|
||||
|
||||
Confidence: medium
|
||||
|
||||
@@ -294,6 +303,8 @@ What is still missing:
|
||||
|
||||
- the transition from any of these branches into a stable active/session state
|
||||
- a deterministic multi-turn reply ladder
|
||||
- evidence that the downstream family frames are actually prompts the host is
|
||||
expected to answer, rather than readable blocks with no immediate follow-up
|
||||
|
||||
## What Feels Stable Right Now
|
||||
|
||||
|
||||
Reference in New Issue
Block a user