Blog
The OneHop blog
Guides, benchmarks, and field notes on Claude, OneHop, and the wider AI model landscape — written to help you ship.
Claude Desktop Still Has No Linux Build: What Linux Developers Can Use Instead
Claude Desktop skips Linux, but Claude Code, MCP, and careful tooling cover most developer workflows without unsafe repackages.
Claude Opus 4.8 Error Spikes: What Developers Should Change After the June 16-19 Incidents
Claude’s June 16-19 error spikes show why production AI apps need retry budgets, fallback routing, and status-aware degradation.
Claude Cowork’s local VM shows the real cost of safe desktop agents
Claude Cowork’s local VM proves agent safety is now an OS/runtime problem, with real UX and IT costs.
Claude Fable 5 vs Opus 4.8 Fast Mode: Same Token Price, Different Trade-Off
Opus 4.8 Fast Mode matches Fable 5’s token price, but speed, access, safeguards, and steering make the choice less obvious.
Run Claude Code safely in a dev container with auto mode and locked-down permissions
A practical guide to isolating Claude Code, restricting egress, disabling bypass mode, and routing Claude Fable 5 through OneHop.
Claude Fable 5 on CursorBench 3.1: 72.9% accuracy on real multi-file coding tasks
Fable 5 leads CursorBench 3.1 at 72.9%, but its price changes the model-choice math for coding agents.
Anthropic’s Opus 4.6 sabotage risk report explains how it monitors internal Claude Code use
Anthropic says it monitors most internal Claude Code use. Developers should copy the control pattern, not the marketing.
Auditing Claude Code Plugins Before You Install the Official Marketplace
A practical checklist for reviewing Claude Code plugins, MCP servers, hooks, permissions, and safer CLI alternatives before install.
Claude Fable 5 on Bedrock and the new 30-day data retention tradeoff
Claude Fable 5 makes Bedrock teams trade frontier capability for 30-day retention and provider data sharing.
Claude Opus 4.8 Fast Mode vs GPT-5.5 Codex Fast Mode for Latency-Bound Coding
A practical price, speed, and context guide for choosing Opus 4.8 Fast Mode or GPT-5.5 Codex Fast Mode in coding loops.
Use Claude Code Review as a CLAUDE.md and REVIEW.md pull request gate
Turn Claude Code Review into a practical PR gate by encoding team rules in CLAUDE.md and REVIEW.md.
How to route Claude Fable 5, Opus 4.8, and Sonnet 4.6 by task and budget
Build a practical Claude routing layer that saves Fable 5 for high-value agents and defaults routine work to Sonnet.
Claude Fable 5 on Terminal-Bench 2.1: 88.0% vs GPT-5.5 at 83.4% and Gemini 3.1 Pro at 70.7%
Fable 5 leads Anthropic’s Terminal-Bench 2.1 table, but price, safeguards, and security-coding results change the decision.
Claude Agent SDK Credit: What Changes for claude -p, GitHub Actions, and Third-Party Agent Tools on June 15
Anthropic is splitting headless Claude workflows into a separate monthly Agent SDK credit. Here is what developers need to change.
Anthropic’s AI-enabled cyber threat report says MITRE ATT&CK misses agentic attack behavior
Anthropic mapped 832 banned cyber accounts to MITRE ATT&CK and found the dangerous part is orchestration, not technique count.
Claude Opus 4.8 vs GPT-5.5 when Fable 5 is unavailable
Fable 5 is offline, so developers need a practical fallback: Claude Opus 4.8, GPT-5.5, or provider routing.
Running a 7-Hour Autonomous Claude Code Task in a vLLM-Scale Codebase
A practical workflow for using Claude Code on huge ML infra tasks without trusting vibes, hidden bugs, or runaway autonomy.
Use Claude prompt caching with cache_control: 5-minute vs 1-hour cache costs
Use Claude prompt caching correctly, compare 5-minute and 1-hour cache writes, and wire Claude Fable 5 through OneHop.
Drop-in Claude API Access Through OneHop
Point your existing Anthropic SDK at OneHop and call Claude Fable 5 with a base URL swap, no client rewrite.
Claude Fable 5 uses classifiers that route some cyber, bio, and distillation requests to Opus 4.8
Anthropic’s Fable 5 routes flagged cyber, bio, chemistry, and distillation requests to Opus 4.8. Developers need to plan for it.
Claude Opus 4.8 and dynamic workflows: why Anthropic is pushing agent orchestration, not just model scores
Opus 4.8 matters less as a benchmark bump than as Anthropic’s bet on long-running, parallel developer agents.
Claude Opus 4.8 on SWE-bench Pro: 69.2% vs GPT-5.5 and Gemini 3.1 Pro
Opus 4.8 leads GPT-5.5 and Gemini 3.1 Pro on SWE-bench Pro, a harder signal for autonomous repo work.
Getting Started with Claude Fable 5: A Practical Developer Guide
Call Claude Fable 5 today with Anthropic’s SDK or OneHop, with current pricing, context limits, and runnable examples.
The US Just Cut Off Foreign Access to Claude Fable 5 and Mythos 5
Anthropic disabled Claude Fable 5 and Mythos 5 after a June 12 US export-control directive. Here’s what developers need to know.
How DXC Is Using Claude for Legacy Code Modernization in Regulated Industries
DXC’s Claude alliance shows how agentic modernization may work in banks, airlines, insurers, factories, and government.
