Skip to main content
株式会社オブライト

Column

Useful articles about SEO, Web Development, and IT

749 articles

AI2026-07-26
Claude Opus 5 Deep Dive — Anthropic's July 24, 2026 Release Keeps $5/$25 While Frontier-Bench Goes 18.7% → 44.4% 1M Context, 128K Output (300K via Batch), the New xhigh Effort Level, OSWorld 2.0 at 70.57% and SWE-bench Verified at 96.0% Opus 4.8 Goes Legacy, Opus 4.1 Retires August 5, 2026 — and thinking: disabled with xhigh Now Returns a 400
Anthropic released Claude Opus 5 on July 24, 2026. Pricing holds at $5 / $25 per M tokens, unchanged from Opus 4.8, while Frontier-Bench v0.1 jumps 18.7% → 44.4% (xhigh), OSWorld 2.0 goes 55.7% → 70.57%, and ARC-AGI-3 goes 1.52% → 30.16%. It ships a 1M context window (default = maximum) with 128K max output (300K via the Batches API), and drops the minimum cacheable prompt to 512 tokens. The new xhigh effort level — above high, below max — targets agentic runs of 30+ minutes and actually beats max (43.3%) on Frontier-Bench. Breaking changes: thinking cannot be disabled at xhigh or max (the API returns a 400), Opus 4.8 is now legacy, and Opus 4.1 retires August 5, 2026. SWE-bench Verified hits 96.0%, but SWE-bench Pro at 79.2% stays just short of Fable 5's 80.0% — Opus 5 replaces Opus 4.8, not Fable 5.
Claude Opus 5AnthropicLLM+4
Business DX2026-07-25
How to Audit Your Company's SaaS Subscriptions — Finding Duplicates and Ghost Accounts
A practical guide for small businesses to audit recurring SaaS costs: how to inventory contracts, spot duplicate tools and ghost accounts, and decide what to cut.
中小企業DXひとり情シスIT予算+2
AI2026-07-25
OpenAI Codex 26.715 — ChatGPT Voice (GPT-Live) and Multi-Folder Projects
Codex app update 26.715 (July 23, 2026): GPT-Live-powered ChatGPT Voice steers work across Chat, Work, and Codex, and local projects gain multi-folder support.
OpenAI CodexCodexOpenAI+3
Business DX2026-07-24
Who Owns the Software You Paid to Build? Source Code and Copyright Explained
Paying for development does not automatically make the copyright or source code yours. This guide explains ownership basics, the contract clauses to check, and what goes wrong when you cannot get the source code.
中小企業DXシステム発注ITリスク対策
AI2026-07-24
GGUF Quantization: Which Level to Pick (Q4_K_M, Q5_K_M, Q8_0, IQ) for Local LLMs
Start with Q4_K_M; step up to Q5_K_M or Q6_K if you have VRAM headroom. This guide explains GGUF naming, the quality/speed/VRAM tradeoffs per level, IQ (imatrix) quants, and how to choose by task. Updated July 2026.
GGUF量子化ローカルLLM+4
Business DX2026-07-23
How SMBs Should Plan Their IT Budget — Benchmarks and a Practical Method
How much should a small business spend on IT? A neutral guide to IT budget benchmarks, cost breakdown, a 5-step planning method, and a checklist for non-technical owners.
中小企業DXシステム発注IT予算+1
AI2026-07-23
Ollama vs LM Studio — How to Choose a Local LLM Runner (2026)
A 2026 comparison of Ollama and LM Studio, the two leading local LLM runners: CLI vs GUI, install and usage, API, OS support, and model formats, with guidance on which to pick.
OllamaLM StudioローカルLLM+3
AI2026-07-23
Our Column Was Featured in Freelance Hub's AI Media Roundup
Our column was featured in Freelance Hub's editorial roundup of media useful for AI-driven development and AI agents. Here's what was highlighted, and a guide to the related articles on this site.
AI中小企業DX
Business DX2026-07-23
Office PC Won't Start or No Internet? A Triage Checklist
A triage checklist for office PCs that won't start and networks that won't connect: 'one person or everyone?' narrows half the causes. Correct router restart order, symptom-based PC steps, stopgap operations, and repair-vs-replace judgment.
IT緊急対応中小企業DXひとり情シス
Business DX2026-07-23
Email Not Sending or Receiving? Triage by Cause
Triage for email that won't send or arrive: narrowing causes by symptom, dealing with spam-flagging (SPF/DKIM in plain terms), the domain-expiry blind spot, and a table of exactly who to contact for each cause.
IT緊急対応中小企業DX保守運用
Business DX2026-07-23
Website Defaced or Down? First Response by Symptom
Site won't load, looks wrong, or triggers a browser warning — here's how to tell them apart and what to do first for each symptom.
IT緊急対応ITリスク対策中小企業DX
AI2026-07-23
GPT-5.6 (Sol / Terra / Luna) vs Grok 4.5 — API Pricing, Specs and How to Choose [July 2026]
GPT-5.6 ships in three tiers — Sol $5/$30, Terra $2.50/$15, Luna $1/$6 (input/output per 1M). xAI Grok 4.5 is $2/$6 with a 500K context and an OpenAI-compatible API. A practical comparison plus per-use-case routing, prompt caching, and reasoning-effort cost tips for SMBs. Updated July 2026.
GPT-5.6Grok 4.5OpenAI+3
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263