Kalidasa–Wordsworth Creative Plugin for Claude Code
❯ “create me a sanskrit verse/poem on vasantha aagamana using pratyabhijna / pce plugin”
⏺ Skill pratyabhijna-creative-engine:poem loaded in Claude Code.
⏺ Sākṣī check passed, then delegated to pratyabhijna-poet via the PCE cascade.
⏺ ★ Insight: the vimarśa (aspect‑shift) detector ran but never fired — the spring → recognition mapping stayed internally consistent across all five śakti‑axes, so no sublation cycle was needed. The surface at candidate index 4 was stable enough to ship as‑is.
वसन्तागमन — Vasantāgamana
(arrival of spring through icchā, jñāna, kriyā, cit, ānanda)
स्तवक १ — इच्छा & ज्ञान (Will & Knowing)
देवनागरी:
▎ वसन्तागमने शिवः शक्तिपञ्चप्रबोधकः।
▎ इच्छाज्ञानमयी भूमि अशोकेन विराजते।।१।।
English:
As spring arrives, Śiva awakens the five‑fold power within;
the earth, suffused with will and knowing, shines in ashoka bloom.
स्तवक २ — क्रिया & चित् (Action & Consciousness)
देवनागरी:
▎ मलयमारुते मन्दे क्रियाशक्तिः समन्विता।
▎ भ्रमरकूजने मधुरे चिद्भूमिः प्रकाशते।।२।।
English:
In the gentle Malaya wind, action‑power unfolds its dance;
in the bees’ honeyed music, the field of consciousness gleams.
स्तवक ३ — आनन्द & प्रत्यभिज्ञा (Bliss & Self‑Recognition)
देवनागरी:
▎ पृथिव्याः जागरणकाले आनन्दशक्तिः वर्धते।
▎ प्रत्यभिज्ञानये काले शिवः स्वमयीं पश्यति।।३।।
English:
In the earth’s awakening, bliss‑power overflows;
in self‑recognition, Śiva beholds himself in every flower.

🔗 Links & Deep Dive
📄 Full mechanism‑study paper (PDF):https://sharathsphd.github.io/pratyabhijna/paper/main.pdf
📊 Overview & results site (PCE v0.4):https://sharathsphd.github.io/pratyabhijna/
💻 Open‑source repo (plugin + CLI + benchmarks):https://github.com/SharathSPhD/pratyabhijna

Your AI coding agent just confidently edited the wrong file. Your research agent cited a retracted paper twice. Your customer-service bot quoted a policy from 2023. None of these are model failures. They're all context failures.Introducing Pratyakṣa, an open-source context engineering harness that fixes how AI agents manage what they know, why they know it, and when to stop trusting it.Here's what makes it genuinely different:🔹 Ancient wisdom meets modern AI — The design vocabulary comes from classical Indian epistemology (Nyāya, Advaita Vedānta, Sāṃkhya). A 14th-century logician's framework for what counts as valid knowledge turned out to be the missing operating manual for AI agents.🔹 Sublation, not deletion — When a newer, more authoritative source supersedes an older one, the old fact is retired, not erased. The audit trail stays intact. Every decision is replayable.🔹 Separation of attention & judgement — Two sub-agents, Manas (selects evidence) and Buddhi (judges and responds), ensure these are never confused — the source of a whole class of hallucinations.🔹 Results that hold up — Across 7 benchmarks (RULER, HELMET, SWE-bench, HaluEval, TruthfulQA and more), the harness beats unaided baselines at p ≤ 0.002. On a 720-pair head-to-head, it anchored on the correct file 720/720 times. The unaided baseline? Coin-flip rate — 50.3%.🔹 Installs in 30 seconds — Works in Cursor, Claude Code, and Claude Desktop. No fine-tuning, no architecture changes. Just two commands.🔌 Plugin (install & use) https://github.com/SharathSPhD/pratyaksha-context-eng-harness📁 Full harness, experiments & validation (GitHub) https://github.com/SharathSPhD/context-engineering-harness📄 Citable preprint — Zenodo (DOI-backed canonical v2.0) https://zenodo.org/records/19653013🚀 v2.1.1 Release page (plugin zip, figures, checksums) https://github.com/SharathSPhD/pratyaksha-context-eng-harness/releases/tag/v2.1.1🗺️ Project status canvas https://sharathsphd.github.io/context-engineering-harness/canvas.html📰 Full article (Substack) https://open.substack.com/pub/technektar/p/when-the-context-window-is-big-and

AI systems that can “imagine” possible futures, called world models, have been applied to Atari games, robotics, and board games. But applying them to economic environments like retail pricing has been largely unexplored. DreamPrice takes an initial step into that space. And solving the fundamental problem that makes economics different from physics turns out to require a detour through a 1978 econometrics paper.