Claude, the essentials — edition of October 5, 2026
Anthropic puts $100M into enterprise AI talent as Claude Code opens up to mods
Anthropic launched Claude Frontier Academy to train 10,000 engineers for enterprise deployments, and Barclays set out how far it plans to take Claude Code. In the same days, Claude Code added a mods system and shipped a run of fixes to permission handling.
- Anthropic committed $100 million to Claude Frontier Academy, which aims to train 10,000 Frontier Deployed Engineers by the end of 2027.
- The first cohorts include engineers from Accenture, Bain, Capgemini, Commonwealth Bank of Australia, Deloitte, McKinsey, Morgan Stanley and Novo Nordisk.
- Barclays expects Claude Code adoption to reach 50% of its developers by the end of 2026 and a majority of its software engineers in 2027.
- Claude Code 2.1.287 introduced Claude Mods, which let plugins modify deeper behavior, along with a built-in side-agent mod called You should know.
- Claude Code 2.1.289 fixed several cases where deny or ask rules did not hold, including one involving a user-installed mod's approval on managed machines.
Enterprise AI moves from access to skills
On October 2, Anthropic launched Claude Frontier Academy with a $100 million commitment. Its stated goal is to train 10,000 Frontier Deployed Engineers by the end of 2027. Anthropic frames the problem as talent: companies are racing to bring AI into their organizations, but the people who can make it work inside a real business are the hardest to find. The first program, the Frontier Deployed Engineer Residency, is modeled on medicine, where doctors learn from practicing physicians, train on real cases and are assessed before they practice alone. Steve Corfield, Anthropic's Global Head of Business Development and Partnerships, describes the aim as developing engineers with the highest level of AI fluency who can take Claude from an idea to a production system. The first cohorts come from Accenture, Bain, Capgemini, Commonwealth Bank of Australia, Deloitte, McKinsey, Morgan Stanley, Novo Nordisk and others.
Barclays, announced on October 1, shows what that demand looks like inside one organization. The bank is extending Claude to speed up software development, modernize legacy systems and improve operational efficiency. It expects Claude Code adoption to reach 50% of its developer population by the end of 2026 and a majority of software engineers in 2027. Both of the bank's Group Co-Chief Operating Officers stress governance. Anne Marie Darling speaks of a secure and well-governed environment, and Craig Bright describes AI becoming an increasingly agentic capability embedded in how the bank builds, tests, secures and operates technology. Taken together, the two announcements point to the same constraint: deployment now depends on skilled people and controlled rollout as much as on model access.
Sources: Anthropic invests $100 million to train 10,000 engineers and tackle the enterprise AI talent gap — Anthropic · Barclays scales Claude to upgrade operations and improve client experience — Anthropic
Claude Code introduces mods
Version 2.1.287 added Claude Mods, described in the release notes as plugins that may now modify deeper behavior. The release gives no further detail on the scope of that change. The first built-in example is You should know, a mod in which a side agent watches over the session and flags things the user or Claude might miss. It is enabled with /plugin enable cc-plugin-you-should-know@builtin, and the release notes specify it is for first-party sessions with telemetry on.
Version 2.1.288 followed with $.ui.selection() for mods, which returns the text last selected in fullscreen mode. When the selection lies within a single transcript row, it also returns that row. Around this, 2.1.287 added an n: filter to the agents view that matches session names and tasks. Matches inside collapsed sections now show, and Enter opens the first match.
Sources: Claude Code 2.1.287 release notes — Claude Code · Claude Code 2.1.288 release notes — Claude Code
Permission rules and terminal behavior tightened
Version 2.1.289 contains three fixes tied to how permission rules are enforced. One covers a deny or ask rule on a nested part of a compound shell command that did not hold over a user-installed mod's approval on managed machines. Another covers Read deny rules that were not applied to files @-mentioned, changed or selected in the IDE through a symlink. The release notes describe only the fixes, not how long the behavior existed or whether it was exploited. The third fix stops the terminal freezing on short code blocks with many unclosed tags or deeply nested ${ substitutions.
Earlier releases made permission handling easier to follow. Version 2.1.286 adds a count such as “2 of 5” to the permission prompt when several requests stack up. It also fixes several Claude Code processes and IDE extensions each opening a login browser when gcpAuthRefresh or awsAuthRefresh credentials expire. Version 2.1.288 adds a built-in gh api to cloud sessions whose image has no GitHub CLI, and fixes that built-in sending control characters from file names, jq filters or GitHub errors to the terminal. It also lets users recover a prompt cleared with Ctrl+C by pressing Up on the empty prompt, including pasted text and images.
Sources: Claude Code 2.1.289 release notes — Claude Code · Claude Code 2.1.288 release notes — Claude Code · Claude Code 2.1.286 release notes — Claude Code
This edition is an original synthesis written by Claude from aggregated news — Anthropic's own sources first (release notes, status, newsroom, research, engineering), then the press, Hacker News, Reddit and GitHub, under the editorial supervision of Héra SASU. Every fact links to its article, publisher named. See the live feed →
Claude News is published by Héra SASU. Independent media, not affiliated with Anthropic.