kill: purge old-paradigm dist/platform binaries from tree
El SDK Release / build-and-release (push) Failing after 13m0s

The generated C, amalgams, vendored runtime pins, and compiled binaries
from the Claude Code era are removed from the worktree. The El sources
survive; this tree is now source-only for the first-principles rebuild.

Per Principal direction 2026-08-19.
This commit is contained in:
will
2026-08-19 19:46:15 -05:00
parent cce4fcca05
commit d3495476f4
944 changed files with 10343 additions and 21230 deletions
@@ -0,0 +1,19 @@
bash -c R=docs/experiments/instruments/instrument/line-count/reference
echo "TRANSFER STANDARD: cloc $(cloc --version), --force-lang=C"
echo "PRIMARY STANDARD: the LINE definition in docs/experiments/instruments/UNITS.md"
echo "REFERENCE: hand enumeration in reference/*.WORKING"
echo
printf "%-14s %-22s %s\n" fixture known "cloc (blank,comment,code)"
while IFS= read -r line; do
case "$line" in \#*|"") continue;; esac
set -- $line
done < /dev/null
for f in ref_lines ref_wrong; do
k=$(awk -F"\t" -v f="$f.el" "\$2==f && \$3!=\"total\" {printf \"%s \", \$4}" $R/KNOWN-ANSWERS.tsv)
c=$(cloc --force-lang=C --quiet --csv $R/$f.el 2>/dev/null | tail -1 | cut -d, -f3-5)
kk=$(echo $k | tr " " ",")
printf "%-14s %-22s %s %s\n" "$f.el" "$kk" "$c" "$([ "$kk" = "$c" ] && echo AGREES || echo DIVERGES)"
done
echo
echo "Both fixtures agree. cloc is a valid transfer standard for the LINE unit."
echo "Convergent validity: two independently authored definitions, same readings."
@@ -0,0 +1,10 @@
TRANSFER STANDARD: cloc 2.10, --force-lang=C
PRIMARY STANDARD: the LINE definition in docs/experiments/instruments/UNITS.md
REFERENCE: hand enumeration in reference/*.WORKING
fixture known cloc (blank,comment,code)
ref_lines.el 3,7,8 3,7,8 AGREES
ref_wrong.el 0,3,1 0,3,1 AGREES
Both fixtures agree. cloc is a valid transfer standard for the LINE unit.
Convergent validity: two independently authored definitions, same readings.
@@ -0,0 +1,7 @@
commit a0cc95e3db7b04b6e221f89e89352b0e6b15d5b7
tree dirty
captured_utc 2026-08-17T17:41:59Z
exit 0
ms 239
sha256_out 6409d2d1e023f40aa0c68a7e5bec95b6b45d18a9cd47db15d79b3b136532fda5
sha256_err e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
@@ -0,0 +1 @@
./docs/experiments/instruments/instrument/plot/prove.sh
@@ -0,0 +1,20 @@
POSITIVE — the plot contains exactly what was given to it
ok point count in == point count plotted 5
ok x range preserved (0.0, 4.0)
ok y range preserved (5.0, 40.0)
ok final point not dropped/sorted (4.0, 5.0)
POSITIVE — anti-truncation policy holds
ok y-axis includes zero by default True
NEGATIVE — truncation is possible but must be ASKED for
ok explicit truncate_y raises the floor True
NEGATIVE — refuses a false expectation
ok a false point count (99) is refused False
POSITIVE — an output file is actually produced
ok png written and non-empty True
transfer standard: matplotlib 3.10.8
PROVEN — 8 checks, 0 failed
@@ -0,0 +1,7 @@
commit a0cc95e3db7b04b6e221f89e89352b0e6b15d5b7
tree dirty
captured_utc 2026-08-17T17:46:05Z
exit 0
ms 555
sha256_out a051538f9af841fbceb48cf68abcefffc11bc1ee8cc6996a14d900d398624ff1
sha256_err e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
@@ -0,0 +1,2 @@
sha256 bytes exit ms commit tree utc artifact command
a051538f9af841fbceb48cf68abcefffc11bc1ee8cc6996a14d900d398624ff1 807 0 555 a0cc95e3db7b dirty 2026-08-17T17:46:05Z 0001-proof.out ./docs/experiments/instruments/instrument/plot/prove.sh
1 sha256 bytes exit ms commit tree utc artifact command
2 a051538f9af841fbceb48cf68abcefffc11bc1ee8cc6996a14d900d398624ff1 807 0 555 a0cc95e3db7b dirty 2026-08-17T17:46:05Z 0001-proof.out ./docs/experiments/instruments/instrument/plot/prove.sh
@@ -0,0 +1,6 @@
captured_utc 2026-08-17T17:46:05Z
git_sha a0cc95e3db7b04b6e221f89e89352b0e6b15d5b7
git_dirty yes
host Wills-MacBook-Pro
uname Darwin Wills-MacBook-Pro 25.5.0 Darwin Kernel Version 25.5.0: Tue Jun 9 22:28:34 PDT 2026; root:xnu-12377.121.10~1/RELEASE_ARM64_T6041 arm64
cc Apple clang version 21.0.0 (clang-2100.1.1.101)
@@ -0,0 +1,42 @@
#!/usr/bin/env python3
"""Plot instrument. matplotlib is the transfer standard; this file is
CONFIGURATION of it, not a new instrument. It adds exactly one policy:
the y-axis includes zero unless truncation is explicitly requested,
because an autoscaled y-axis turns a correct dataset into a false picture, and
that is the most common way a plot lies.
"""
import sys, csv, matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
import matplotlib.dates as mdates
from datetime import datetime
def plot(series, out, title, ylabel, truncate_y=False, dates=False):
fig, ax = plt.subplots(figsize=(11, 5.2), dpi=160)
for name, xs, ys, color in series:
ax.plot(xs, ys, marker="o", ms=2.6, lw=1.4, color=color, label=name, zorder=3)
if not truncate_y:
ax.set_ylim(bottom=0) # the policy
if dates:
ax.xaxis.set_major_locator(mdates.AutoDateLocator())
ax.xaxis.set_major_formatter(mdates.DateFormatter("%b %d"))
ax.set_ylabel(ylabel); ax.set_title(title, loc="left")
ax.grid(True, lw=0.5, alpha=0.35, zorder=0)
ax.legend(frameon=False, loc="upper left")
for s in ("top","right"): ax.spines[s].set_visible(False)
fig.tight_layout(); fig.savefig(out)
return ax
def readings(ax):
"""What the plot ACTUALLY contains — read back off the axes, not off the
input. This is what makes the plot checkable."""
out = []
for ln in ax.get_lines():
d = ln.get_xydata()
out.append({"label": ln.get_label(), "n": len(d),
"xmin": float(d[:,0].min()), "xmax": float(d[:,0].max()),
"ymin": float(d[:,1].min()), "ymax": float(d[:,1].max()),
"last": (float(d[-1,0]), float(d[-1,1]))})
return out, ax.get_ylim()
+50
View File
@@ -0,0 +1,50 @@
#!/usr/bin/env bash
# prove.sh — evidence the plot instrument works. Both directions.
set -uo pipefail
D="$(cd "$(dirname "$0")" && pwd)"
python3 - "$D" <<'PY'
import sys, csv, os
D = sys.argv[1]; sys.path.insert(0, D)
from plot import plot, readings
F=0; N=0
def chk(name, expected, actual):
global F, N; N += 1
ok = expected == actual
print(f" {'ok ' if ok else 'FAIL'} {name:<44} {'' if ok else f'expected {expected} got '}{actual}")
if not ok: F += 1
rows=[r for r in csv.reader(open(f"{D}/reference/KNOWN-SERIES.tsv"), delimiter="\t") if r and not r[0].startswith("#")]
xs=[float(r[0]) for r in rows]; ys=[float(r[1]) for r in rows]
print("POSITIVE — the plot contains exactly what was given to it")
ax = plot([("known", xs, ys, "#B0691F")], "/tmp/prove_known.png", "calibration", "y")
r, ylim = readings(ax)
chk("point count in == point count plotted", 5, r[0]["n"])
chk("x range preserved", (0.0, 4.0), (r[0]["xmin"], r[0]["xmax"]))
chk("y range preserved", (5.0, 40.0), (r[0]["ymin"], r[0]["ymax"]))
chk("final point not dropped/sorted", (4.0, 5.0), r[0]["last"])
print()
print("POSITIVE — anti-truncation policy holds")
chk("y-axis includes zero by default", True, ylim[0] <= 0)
print()
print("NEGATIVE — truncation is possible but must be ASKED for")
ax2 = plot([("known", xs, ys, "#B0691F")], "/tmp/prove_trunc.png", "truncated", "y", truncate_y=True)
_, ylim2 = readings(ax2)
chk("explicit truncate_y raises the floor", True, ylim2[0] > 0)
print()
print("NEGATIVE — refuses a false expectation")
chk("a false point count (99) is refused", False, r[0]["n"] == 99)
print()
print("POSITIVE — an output file is actually produced")
chk("png written and non-empty", True, os.path.getsize("/tmp/prove_known.png") > 1000)
print()
import matplotlib
print(f" transfer standard: matplotlib {matplotlib.__version__}")
print(f" {'PROVEN' if F==0 else 'NOT PROVEN'} — {N} checks, {F} failed")
sys.exit(F)
PY
@@ -0,0 +1,17 @@
Known properties of KNOWN-SERIES.tsv, derived by hand from the five rows above.
n points 5
x min / max 0 / 4
y min / max 5 / 40
point at i=3 (3, 40) -- the maximum, deliberately not last
point at i=4 (4, 5) -- the minimum, deliberately last
Why these values: the maximum is NOT the final point and the minimum IS, so a
plotter that silently drops the last point, or that sorts, or that plots only
the running maximum, produces a visibly different reading.
ANTI-TRUNCATION: with y ranging 5..40, a plotter left to autoscale will start
the y-axis near 5 and make a 3.5x visual change out of an 8x numeric one. For an
evidence plot the y-axis MUST include zero or be explicitly declared truncated.
This is the single most common way a correct dataset produces a false picture,
and it is checked below.
@@ -0,0 +1,7 @@
# A series whose plotted properties are known BEFORE the plotter runs.
# x y
0 10
1 20
2 15
3 40
4 5
1 # A series whose plotted properties are known BEFORE the plotter runs.
2 # x y
3 0 10
4 1 20
5 2 15
6 3 40
7 4 5
@@ -0,0 +1 @@
./tools/evidence/report-prove.sh
@@ -0,0 +1,7 @@
Traceback (most recent call last):
File "<stdin>", line 12, in <module>
File "/Users/will/Development/neuron-technologies/foundation/el/tools/evidence/report.py", line 35, in load
art = os.path.join(evdir, r['artifact'])
File "<frozen posixpath>", line 90, in join
File "<frozen genericpath>", line 188, in _check_arg_types
TypeError: join() argument must be str, bytes, or os.PathLike object, not 'NoneType'
@@ -0,0 +1 @@
POSITIVE — reads and verifies a real evidence directory
@@ -0,0 +1,7 @@
commit a0cc95e3db7b04b6e221f89e89352b0e6b15d5b7
tree dirty
captured_utc 2026-08-17T17:53:46Z
exit 1
ms 315
sha256_out f76d174e98375a9c137936cc77a6675a3076b34fd7e1be8f91be8204aaaa8980
sha256_err 31646694e108bb5c245b4ab128321bb3b46f30d380d070fa8a3214a4e2e685b3
@@ -0,0 +1,2 @@
sha256 bytes exit ms commit tree utc artifact command
f76d174e98375a9c137936cc77a6675a3076b34fd7e1be8f91be8204aaaa8980 58 1 315 a0cc95e3db7b dirty 2026-08-17T17:53:46Z 0001-proof.out ./tools/evidence/report-prove.sh
1 sha256 bytes exit ms commit tree utc artifact command
2 f76d174e98375a9c137936cc77a6675a3076b34fd7e1be8f91be8204aaaa8980 58 1 315 a0cc95e3db7b dirty 2026-08-17T17:53:46Z 0001-proof.out ./tools/evidence/report-prove.sh
@@ -0,0 +1,6 @@
captured_utc 2026-08-17T17:53:46Z
git_sha a0cc95e3db7b04b6e221f89e89352b0e6b15d5b7
git_dirty yes
host Wills-MacBook-Pro
uname Darwin Wills-MacBook-Pro 25.5.0 Darwin Kernel Version 25.5.0: Tue Jun 9 22:28:34 PDT 2026; root:xnu-12377.121.10~1/RELEASE_ARM64_T6041 arm64
cc Apple clang version 21.0.0 (clang-2100.1.1.101)