---
title: 'In the News: August 31, 2026'
description: 'A controlled harness study finds structural constraints alone can hurt agent reproducibility until the plan step itself is schema validated.'
canonical_url: 'https://darkfactory.dev/news/2026-08-31-morning'
markdown_url: 'https://darkfactory.dev/news/2026-08-31-morning.md'
collection: news
date_published: '2026-08-31T07:35:00-04:00'
date_modified: '2026-08-31T07:35:00-04:00'
---

# In the News: August 31, 2026


A deterministic harness did not automatically make two coding agents more reproducible. In two of four test cases it made reproducibility worse until the planning step was schema validated. Anthropic also needed two posts to explain how Claude Code's weekly limits will change.

## 1. Schema-validated planning sharply improves reproducibility in a deterministic harness

**[Harness Engineering for Predictable Agentic Systems](https://arxiv.org/abs/2608.26197)** · Saransh Dhage, independent researcher · arXiv, August 25, 2026

Dhage wrapped two open models, Qwen-2.5-7B and Gemma-3-27B, in a harness of finite-state execution, forced tool selection, output validation and bounded retry, then ran each against two synthetic pipelines 100 times per condition. Against a plain, unconstrained baseline, the harness significantly improved exact-match reproducibility in one of the four model-task combinations, significantly degraded it in two, and had no measurable effect on the fourth, even though task success held steady at 97 to 100 percent throughout. Tracing the result layer by layer showed why: tool order, state sequence and output were already close to fully consistent, so the model's free-text planning step, left unconstrained, became the only place variance could still hide. Adding a schema-validated planning gate before any tool call, which Dhage calls Structured Planning, closed the gap: three of the four cells reached a perfect Reproducibility Rate of 1.000 at 100 runs, the fourth reached 0.980 on account of a small number of genuine tool-calling failures unrelated to plan wording, task success rose to 100 percent in three of the four, and token cost fell 15 to 17 percent versus baseline across the board. The fix was not free for every model. Structured Planning made Qwen 12 to 21 percent faster than the plain harness, and Gemma 15 to 24 percent slower, a split that grew rather than shrank when the sample size was doubled to check it wasn't noise.

**Why it matters:** Finite-state execution and forced tool selection left free-text planning as the main place variance could hide. That made the harness score worse than the unconstrained baseline in two tests. Schema validation cost little in tokens, but its effect on speed depended on which model sat underneath it, so teams should measure that tradeoff before shipping.

## 2. Anthropic needed a second post to say plainly that Claude Code's weekly limits are dropping

**[Anthropic (@ClaudeDevs)](https://x.com/ClaudeDevs/status/2093742321473065266)** · X, August 29, 2026

Starting September 14, Anthropic is permanently raising Claude Code's standard weekly usage limits by 25 percent above the original baseline for Pro, Max, Team and seat-based Enterprise plans. That replaces the temporary 50 percent boost the company has run since its May to August promotion. Measured against what subscribers have right now rather than against the original baseline, the change is a cut: index the original allowance at 100, today's boosted level sits at 150, and the September 14 level lands at 125, a 17 percent reduction from today. Anthropic's first post led with the 25 percent figure. After users worked out the comparison publicly, as [BleepingComputer reported](https://www.bleepingcomputer.com/news/artificial-intelligence/anthropic-is-cutting-claude-codes-current-weekly-limits-by-17-percent/), the company deleted that post and reposted, stating plainly: "Compared to today, this works out to a 17% reduction in weekly limits on Claude Code."

**Why it matters:** Anyone budgeting Claude Code capacity against this summer's boosted limits should plan for roughly five-sixths of that headroom from September 14. Starting from current capacity, the change is a 17 percent cut; the 25 percent increase applies only to the original baseline.