From 5ab9dcf3ac669705a3d1870e414c1c27f7fde94b Mon Sep 17 00:00:00 2001 From: PBNZ Date: Mon, 20 Apr 2026 17:09:56 +1200 Subject: [PATCH] docs(landing): mention v0.3.0 current-gen calibration and link router example MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Adds a short "Calibrated for current-generation models" section to the landing page describing v0.3.0's model-awareness work (outcome-oriented tool guidance, effort-awareness, session memory, parallel dispatch, vision-as-primary-evidence) and pointing at the optional router CLAUDE.md example under docs/examples/. Also refreshes the "Future development" bullets: the roadmap now reflects v0.3.0 as shipped, notes the translation issue is still open, and re-slots sub-agent decomposition to a later release (the v0.3.0 copy called it out as "v0.3.0" which is now misleading). Pure copy/structural update — no JS or CSS changes, OG cards and canonical URL untouched. Co-Authored-By: Claude Opus 4.7 (1M context) --- docs/index.html | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) diff --git a/docs/index.html b/docs/index.html index f9d7669..b1221b3 100644 --- a/docs/index.html +++ b/docs/index.html @@ -218,11 +218,15 @@

What makes Newton different from just prompting “be more critical&rdqu

“Be more critical” produces performed scepticism — manufactured objections that sound rigorous but don't change the answer. Newton is calibrated the other way: push back when there's a real reason, agree plainly when you're right, and flag genuine uncertainty rather than either hedging or arguing to seem sharp. The skill spells out what “real reason” looks like, what self-critique-before-delivery means in practice, and when clarifying questions are load-bearing versus ceremonial — so the behaviour is consistent turn-to-turn rather than drifting with each prompt.

Newton also operationalises reuse-before-rebuild, current-sources research, surgical editing, and handoff-when-scope-outgrows-the-chat as explicit sub-routines the agent runs, not as vibes. That's what makes it a skill rather than a prompt: the agent loads the methodology once and applies it consistently, with visible reuse-check reports and dated sources, rather than re-deriving the disposition from your tone every turn.

+

Calibrated for current-generation models

+

Newton is model-agnostic by design — the same methodology runs on Opus, Sonnet, and Haiku — but v0.3.0 is tuned so current-generation capabilities (adaptive thinking, implicit tool detection, session memory, parallel execution, high-resolution vision) are leveraged rather than fought. Tool guidance is outcome-oriented instead of prescriptive, independent sub-tasks parallelise, multi-turn sessions can use file-system memory, and visual input is treated as primary evidence. Older models still get the full Newton experience with graceful degradation — no feature relies on a specific model.

+

Want model-specific behaviour layered on top of every skill, not just Newton? v0.3.0 ships an optional router CLAUDE.md example you can drop into your own ~/.claude/CLAUDE.md. It's opt-in — the plugin does not install it for you.

+

Future development

-

Planned after v0.2.0:

+

Planned after v0.3.0: