---
title: 'In the News: October 27, 2026 (Extra 2)'
description: 'One MCP integration carried about 69% of an internal agent''s 177,444-token "hi", and OpenCode''s Code Mode cut that to 13,274.'
canonical_url: 'https://darkfactory.dev/news/2026-10-27-extra-2'
markdown_url: 'https://darkfactory.dev/news/2026-10-27-extra-2.md'
collection: news
date_published: '2026-10-27T16:20:00-04:00'
date_modified: '2026-10-27T16:20:00-04:00'
---

# In the News: October 27, 2026 (Extra 2)


An engineer at Bold.org traced runaway token usage in an internal agent to a single MCP server's tool definitions, and turned on an experimental OpenCode mode to fix it. A separate post, at 600 points on Hacker News when we looked, uses one engineer's GitHub history to argue that cheap implementation raised his workload instead of lowering it.

## 1. One MCP server was 69% of an agent's starting context

**[Why Saying 'Hi' to Our OpenCode Agent Used 177,000 Tokens](https://oioannou.com/blog/how-a-hi-can-be-expensive/)** · Orestis Ioannou, VP of Technology at Bold.org · personal blog, September 18, 2026

Ioannou runs an internal company agent on OpenCode 1.18.31 (his deployment, not a benchmark). Token use was climbing faster than sessions or automations could explain. One conversation recorded 50.4 million tokens across 121 model calls, 49.9 million of them cache reads, which first suggested long conversations. Following sessions back to their first call showed an investigation starting at about 177K input tokens before it had gathered anything, and a fresh conversation answering "hi" showed 177,444.

The OpenCode TokenScope report put the skill catalog at about 3,400 tokens and the tool definitions it could inspect at about 5,800, which did not account for the request. It omitted the full MCP definitions. Inspecting those separately, he found that Meta Ads alone had 97 tools and roughly 595K characters of descriptions and parameter schemas, about 72% of the measured MCP definition text. With Meta Ads disabled, the same greeting dropped to 54,181 tokens. With it restored and `OPENCODE_EXPERIMENTAL_CODE_MODE=true` set, the greeting used 13,274, which he reports as about 92% smaller than the original.

He is explicit about the limit of that number: it is the greeting's recorded usage, "not a claim about our final bill." A real investigation still loads definitions and results as it works. He also notes that OpenCode 2 makes Code Mode the default for MCP servers, and that Claude Code's MCP Tool Search and Codex's deferred MCP discovery follow the same on-demand pattern. All figures are his own, from his hosted deployment, and he says an AI assistant helped reconstruct the investigation from session history.

**Why it matters:** If your agent has a long list of integrations, the cost of making every tool available on every request can dwarf the cost of using them, and a dashboard total will not show it. Check the first-call input size of a fresh session before blaming long conversations or reasoning effort.

## 2. Cheap implementation raised one engineer's workload

**[AI Promised Less Work. I Have Never Worked More.](https://oioannou.com/blog/ai-promised-less-work/)** · Orestis Ioannou, Bold.org · personal blog, October 27, 2026

Ioannou pulled his GitHub contribution counts, private work included, for every month from January 2019. As Principal Engineer he averaged 114 contributions a month from August 2022 to July 2025, against 173 as a full-stack engineer. By October 27, 2026 he was at 2,102, more than any full year from 2021 to 2025. August 2026 was the busiest month in the timeline at 427, against a previous high of 319 in April 2020. His median weekday went from 3 or 4 contributions (2022 to 2025) to 7, and he counts 16 days with 24 or more contributions this year, against one per year before.

He calls the counts "a rough signal of activity, not a measure of value," and they mix commits, pull requests, reviews and issues. This is one person's record, with an incomplete 2019 baseline. His account of the cause is that running several agents in parallel means each one returns with a question, a plan or a diff, and the context switching is what tires him. He describes three bursts since January, each followed by a pullback.

**Why it matters:** This is a single self-reported case, not a measurement of agent productivity. It does name a constraint that is easy to miss when planning agent fleets: the human attention needed to review and decide scales with the number of agents running, not with how cheap each task has become.