Child of #205. Cold read of both sites, 2026-09-03.
#205 diagnoses this site as too verbose and asks for cognitive-load reduction. robertdelanghe.dev already solved the same problem for the same author, and the solution is structural rather than editorial. Worth copying the structure, not the words.
What the sibling does
Every entry in "Selected work" is a four-row definition list with the label in a mono small-caps gutter and the content in a serif measure:
PROBLEM Agent-authored commits reached main without review, and afterwards
nobody could say which writes the agent had been authorized to make.
INTERVENTION Moved privileged git effects behind capability-checked operations,
and routed the work through one pipeline whose only exit is a merged
pull request.
EVIDENCE prx drives a work unit to an attributed pull request; ocap-provenance
records which capabilities each effect held, and which were denied.
Both are public, and both name what is still outside the check.
ROLE System model, schemas, implementation, and the operational testing.
The schema is doing the work. You cannot write "the bet" into a PROBLEM row, or "convergence is in progress" into an EVIDENCE row. The form rejects the register that #205 objects to. Compare the third entry, which lands a number and its instrument in one breath:
EVIDENCE — Critical bugs fell 20% — critical-severity tickets per quarter, the four quarters before against the four after.
That is the house epistemics — claim, then the measurement behind it — expressed as a layout rather than as a policy.
The sharpest observation
The Bounded Systems row on the personal site explains Bounded Systems better than bounded.tools does. Four rows, roughly 70 words, and a stranger comes away knowing what broke, what was done, and what is still uncovered. bounded.tools spends a hero, a demo block, three cards and a graded-status band getting somewhere less concrete.
The origin story in that PROBLEM row — agent commits reaching main, and nobody able to say afterwards what the agent had been authorized to do — does not appear anywhere on bounded.tools. It is the best sentence either site has about why this project exists, and it is on the wrong domain.
Measured differences in how the two pages are built
|
robertdelanghe.dev |
bounded.tools |
| Page height, desktop |
2,928 px |
3,623 px |
| Content column |
880 px, single measure |
alternating full-bleed bands |
| Type roles |
3, consistently assigned (mono labels, serif prose, sans headings) |
mixed |
| Colour |
one accent, on the item numbers |
large green and black section blocks |
| Whole career on one page |
yes |
n/a |
The sibling fits more information into less height because hairline rules and a gutter carry the hierarchy. This site spends vertical space on coloured bands that signal "new section" and little else. That is a straight trade of scroll depth for decoration.
Suggested shape
- Rebuild "What's here" as the four-row grammar.
guest-room, claude-box and prx each have a real problem, a real intervention, and public evidence; the current cards give a tagline and a paragraph.
- Move the origin story onto the home page, in the
PROBLEM row where it belongs.
- Drop the full-bleed bands for rules and a single measure, and let the accent colour mark items rather than regions.
- Make the
EVIDENCE row link to the code, as the sibling does inline. That is the same discipline /conformance applies to claims, applied to the landing page.
Note the honest limit: the sibling also fails cognitive.coga-obj5, so this will not automatically clear that gate. The argument for it is legibility to a stranger, not a green row.
Filed from a cold read, unclaimed and unassigned. Related: #205, #273, #274.
Child of #205. Cold read of both sites, 2026-09-03.
#205 diagnoses this site as too verbose and asks for cognitive-load reduction. robertdelanghe.dev already solved the same problem for the same author, and the solution is structural rather than editorial. Worth copying the structure, not the words.
What the sibling does
Every entry in "Selected work" is a four-row definition list with the label in a mono small-caps gutter and the content in a serif measure:
The schema is doing the work. You cannot write "the bet" into a
PROBLEMrow, or "convergence is in progress" into anEVIDENCErow. The form rejects the register that #205 objects to. Compare the third entry, which lands a number and its instrument in one breath:That is the house epistemics — claim, then the measurement behind it — expressed as a layout rather than as a policy.
The sharpest observation
The Bounded Systems row on the personal site explains Bounded Systems better than bounded.tools does. Four rows, roughly 70 words, and a stranger comes away knowing what broke, what was done, and what is still uncovered. bounded.tools spends a hero, a demo block, three cards and a graded-status band getting somewhere less concrete.
The origin story in that
PROBLEMrow — agent commits reaching main, and nobody able to say afterwards what the agent had been authorized to do — does not appear anywhere on bounded.tools. It is the best sentence either site has about why this project exists, and it is on the wrong domain.Measured differences in how the two pages are built
The sibling fits more information into less height because hairline rules and a gutter carry the hierarchy. This site spends vertical space on coloured bands that signal "new section" and little else. That is a straight trade of scroll depth for decoration.
Suggested shape
guest-room,claude-boxandprxeach have a real problem, a real intervention, and public evidence; the current cards give a tagline and a paragraph.PROBLEMrow where it belongs.EVIDENCErow link to the code, as the sibling does inline. That is the same discipline/conformanceapplies to claims, applied to the landing page.Note the honest limit: the sibling also fails
cognitive.coga-obj5, so this will not automatically clear that gate. The argument for it is legibility to a stranger, not a green row.Filed from a cold read, unclaimed and unassigned. Related: #205, #273, #274.