Think with AI
& not just about it.
Choose better AI tools. Write better prompts. Build smarter workflows.
The model is not the system. Start with the free chapter, then install an OS.
Welcome to T-Minus AI
I'm Kapisch, an AI Adoption Architect & Solo Builder.
I help leadership teams turn scattered AI tools, pilots, and experimentation into role-specific workflows, measurable adoption, and internal capability. At OneKapisch, I build focused, privacy-first software for macOS, iOS, Android, and the web. No unnecessary accounts, tracking, or bloat. T-Minus AI is where I publish the research and workflows behind better AI decisions.
Get the Weekly AI Signal
One email every Tuesday for operators and builders: the 3 AI moves that matter, what they mean for real workflows, and one practical action to try.
New issue every Tuesday. No spam. Unsubscribe any time.
T-Minus Signal / What's New in AI
The AI landscape, distilled.
Curated signals from the frontier. Newest entries source-checked September 28, 2026.
Claude Sonnet 5 Keeps $2 / $10 as Its Standard Price
Anthropic's pricing page now lists Sonnet 5's $2 input and $10 output per million tokens, first announced as introductory pricing through August 31, 2026, as the standard price. The scheduled September 1 increase to $3 / $15 will not occur.
Why it matters: Any budget built on the planned $3 / $15 rate overstates Sonnet 5 token cost by 50%. Recheck cost models that assumed the increase.
Claude Opus 5.5 Becomes Anthropic's Recommended Starting Model
Anthropic released Claude Opus 5.5 on September 22, 2026 at $4 input and $20 output per million tokens, with a 1M-token context window and up to 128K output tokens. It is available on Pro, Max, Team, Enterprise, and the API.
Why it matters: Anthropic says Opus 5.5 costs 40% less to run than Opus 5 and is over 30% faster at default settings. Treat those as vendor claims and rerun your own evals before moving a production workflow. Claude Fable 5.1 ($10 / $50) remains the higher tier for demanding reasoning and long-horizon agent work.
GPT-6 Sol and GPT-6 Luna Split OpenAI's New Lineup by Cost
OpenAI released GPT-6 Sol ($2 input / $10 output per million tokens) for complex coding and agentic work and GPT-6 Luna ($0.10 / $0.50) for efficient, high-volume work on September 22, 2026. Both list a 1.05M-token context window.
Why it matters: In ChatGPT, OpenAI says GPT-5.5 retires on October 14, 2026 on all plans. GPT-6 Sol replaces it on Plus, Pro, Business, Enterprise, and Edu; GPT-6 Luna replaces it on Free and Go.
GPT-6 Astra Arrives as OpenAI's Most Capable Model
OpenAI released GPT-6 Astra on September 3, 2026 and describes it as its most capable model. API pricing is $10 input and $50 output per million tokens, with a 1.05M-token context window and up to 128K output tokens.
Why it matters: Astra is priced like Claude Fable 5.1 at $10 / $50. In ChatGPT, OpenAI lists Astra in ChatGPT Work and in Codex in the desktop app, so check the models page before assuming it is in your plan's chat picker.
Latest from the blog
Source-dated AI analysis
Reasoning Effort, Explained: Why "Always Use Max" Is Costing You (and Sometimes Hurting Your Answers)
Most people leave AI on maximum reasoning because it feels like the safe choice. It is not. Max effort burns far more tokens, adds latency, and on simple tasks can actually make answers worse. Here is how to choose effort like a pro.
GitHub Copilot's New Usage-Based Billing (June 2026): What It Actually Costs You
On June 1, 2026, GitHub Copilot replaced flat premium-request limits with usage-based AI Credits. Here is what each plan actually includes now, and how to keep the new model from surprising you.
ChatGPT API vs ChatGPT Plus: Which Is Actually Cheaper for You? (2026)
The $20 ChatGPT Plus subscription and pay-as-you-go API pricing answer different questions. The break-even is around 1,379 messages a month, here is how to find which side of it you are on.
How to Forecast Your Monthly AI Spend Before the Bill Arrives (2026)
Most people find out what they spent on AI when the invoice lands. Forecasting flips that, here is the burn-rate method and the tooling that predicts your month-end bill while you can still act on it.
Claude Code Rate Limit Hit? What's Actually Happening and How to Fix It (2026)
Claude Code has two limits running at once, a 5-hour rolling window and a weekly cap. Here is how each one works, what changed in May 2026, and how to stop hitting them mid-task.
Claude API Pricing 2026: Opus 5.5 vs Fable 5.1 vs Sonnet 5 vs Haiku 4.5
Claude API pricing looks simple until the monthly bill arrives. Here are the current Opus 5.5, Fable 5.1, Sonnet 5, and Haiku 4.5 rates, a worked monthly example, and the levers that cut spend the most.
AI Tools
Find the right tool for the job
Building with agents? Read the stack primer before you add another MCP server.Free Guides
Practical playbooks you can use today
Start here
The model is not the system
Where rules live in 2026. ChatGPT Projects are workspaces. Repos need a harness. Eight minutes.
Read the free chapterThe GPT-5 Intelligence Playbook
Professional model selection, execution modes, reasoning escalation, and agentic workflow blueprints for the GPT-5 family.
The OpenClaw Field Manual
A practical manual for OpenClaw architecture, skill risk management, 25+ real-world use cases, and a hardened deployment runbook.
Multimodal AI Mastery
Vision, voice, video, and document workflows with multimodal model comparisons, prompt frameworks, and failure-mode fixes.
AI Stack Builder
Answer a few questions and get the right AI stack for your workflow and budget.
Transcript Summarizer
Paste any transcript — YouTube, podcast, meeting — get briefs, action items, chapters, and prompts.
Prompt Scorecard
Grade any AI prompt in seconds — 7 dimensions, letter grade, fix suggestions.
AI Plan Comparator
Compare real AI plan tiers side-by-side before you stack more subscriptions.
From the same studio
Tools behind the work
T-Minus AI helps you choose the system. These two private Mac tools support the work once the session starts.
Measure the session
Keep usage, spend, subscriptions, and rate-limit pressure visible while you build.
macOS · Local-first · Free + one-time Pro
Build the context first
Prepare ordered, reviewable context before it leaves your Mac.
macOS 14+ · Local-only history · Free + one-time Pro
Trusted by power users
"Perfect for beginners who finally decided to use AI efficiently."
— Halil Önder Özbaşak
"An excellent guide with checklists to get the most accurate prompt and the most relevant result."
— Verified Gumroad buyer
"This power user guide helps build a system for the most appropriate AI answers."
— Verified Gumroad buyer