Paper
A public digest of research, technology discussion, and finance-learning notes. It is written in third person and intentionally excludes private voice notes, social-media summaries, operational state, and personal feedback transcripts.
Latest source mix: 3 September 2026
They found: Two pilot studies with 12 advanced LLM users report roughly halved perceived workload and high usability, suggesting that conversation operations can reduce fatigue beyond model-quality gains.
They found: Eight blind participants issued 1,258 commands across 12 applications; GPT-5 achieved the highest reported success rate at 52.5%, with failures spanning grounding, planning, constraint tracking, and termination.
They found: Sixteen participants used the probe for one week and valued planned roles, timely initiative, lightweight visual representations, and non-directive framing.
They found: AR improved perception-level awareness of the robot but did not improve higher-level robot or environmental awareness; freed attention was redirected to conformal visuals rather than the physical scene.
They found: Across 39 participants, complex overlays, unlabeled tools, and immediate feature density caused trial-and-error behavior; the authors propose parameter transparency, point-of-use guidance, and phased disclosure.
Finance — 2026-09-03 New Australian economy grew 0.4% in the June quarter Source: Australian Bureau of Statistics · 2026-09-02 - Real GDP rose 0.4% quarter on quarter and 2.1% through the year. - Nominal GDP rose 0.8%, while the terms of trade fell 1.6%. - The household saving-to-income ratio edged up to 6.5% from 6.4%. - Private demand and net trade contributed to growth, while inventories detracted. Tutor: Separate real activity from nominal growth: pric…
Vanja reads a private daily edition, talks back to it, and promotes the parts that remain professionally useful into this public page. The private reader and recording tools are intentionally not linked here.
Read the blog posts