The AI space moves fast — and the tool names alone can cause confusion.
Claude vs. Claude Code vs. Claude Cowork. Perplexity vs. Perplexity Computer Use. Gemini vs. Google AI Studio vs. NotebookLM.
This guide cuts through the noise. Whether you're brand new to AI or already a daily user, use this as your go-to reference for picking the right tool for the job.
Click any tool name, card or bubble in this guide to pull up more information.
New & Notable
This week's updates first, then earlier July changes. Anything older than a month rolls off — the durable details live in each tool's card below.
›
New & Notable
This week's updates first, then earlier July changes. Anything older than a month rolls off — the durable details live in each tool's card below.
📅 Updated August 18, 2026
🆕 This Week — August 11–18just added
Grok 4.6 launches — built to stay with long, multi-step tasks
xAI released Grok 4.6 — an upgraded flagship that's especially good at long, multi-step jobs like research, working across a codebase, or turning an idea into a finished app. For everyday users, nothing changes: it's included in your current Grok plan. Builders, take note: it handles a bigger context window (500K tokens ≈ a small library of documents), adds a deeper "xhigh" thinking level, and its per-token API price stays among the lowest of the big labs (exact rates in the Grok card). Live today in the xAI API, Cursor, and Grok Build.
Gemini 3.7 Flash ships — smarter, faster, and half the price to build with
Google launched Gemini 3.7 Flash — its smartest "workhorse" model yet for fast answers, coding, and agent-style tasks, arriving just three weeks after the previous version. Good news: it costs about half as much per request as its predecessor through December 31, 2026 — a rounding error for normal users (your AI Pro plan doesn't change), but a real saving if you build with the Gemini API. Exact rates live in the Gemini card. Rolling out to Spark (AI Pro & Ultra) with better Google Workspace tool use.
Replit adds MCP access, workspace regions, and automatic security scanning
Replit shipped a wave of builder updates. A native Replit MCP Server now connects Replit to ChatGPT, Claude, or Slack so you can create, inspect, and publish apps from inside another tool. Pro builders can pick a workspace region (NA, EU, Asia), and there's a path to migrate Replit Auth apps to Clerk Auth with SSO via Okta/Entra ID (free through Oct 1, 2026). Agent now runs an automatic Semgrep security scan on the files it changes before finishing a task (Aug 7), and Enterprise admins get a beta Admin API for usage and budget reporting.
Base44 adds Branches — build on the side, merge when you're ready
Base44 introduced Branches, a native way to build and experiment on a separate version of your app — a bold redesign, a complex feature, or parallel team work — while your main version stays untouched until you merge. Branches run in parallel (up to five builds at once), draw from the same credit balance as main, and have no limit on how many you keep open. Data, connections, automations, and app settings stay shared across branches. Available on the Builder plan and higher.
Earlier in August — August 4–11kept for the month
Claude Sonnet 5's lower price is locked in — the September increase is off
Anthropic confirmed on August 10 that Claude Sonnet 5 keeps its lower launch price — the increase scheduled for September 1 is canceled. What it means for you: steady, predictable costs, whether you chat on the $20/mo Pro plan or pay per token through the API. Sonnet 5 remains Anthropic's best everyday value and undercuts OpenAI's balanced GPT-5.6 Terra. Exact rates live in the Claude card.
ChatGPT free users get unlimited text chats — plus GPT-5.6 Luna and a "Think" button
OpenAI made GPT-5.6 Luna the default model for Free and Go users, with unlimited text chats rolling out next week (limits still apply to file uploads, images, and other tools). Free/Go users also get a new Think button to access higher reasoning on harder questions. For Plus and Pro, GPT-5.6 Sol in Chat was updated to be more reliable with facts and give more focused answers, plus a new thinking slider to choose how much reasoning each response gets. Separately (Aug 10): ChatGPT Business Premium seats launched at $125/user/mo ($100 annual) — 5× the usage of Standard seats ($25) with no five-hour usage limit, for power users in a Business workspace.
Perplexity Computer gets shared "Projects" workspaces and expanded Model Council
Perplexity upgraded its Computer platform with Projects — shared workspaces that integrate files and use Brain memory to keep context across sessions, plus a Windows app and an overhauled Model Council that lets you build a board of 2–8 models (OpenAI, Anthropic, Google, plus open-weight GLM and Kimi) to tackle one question and synthesize where they agree or conflict. Model Council now reaches individual Pro ($20/mo) users, not just Max. Also: GPT-5.6 Terra and Luna are now live in Perplexity Computer (Terra is the default subagent; Luna runs scheduled automations).
Late Julykept for the month
Perplexity's Personal Computer expands to Windows — and now has "Brain" memory
Perplexity's agentic Personal Computer tool now runs on Windows (joining the Mac version from April), so it can act across local files, Microsoft 365, and the web in one place. Rolling out to Max & Enterprise Max subscribers ($200/mo+). Earlier in July, Perplexity also shipped Brain — self-improving memory that builds a private context graph across your sessions and connectors, refreshing overnight. In testing it lifted answer correctness ~25% and recall ~16%.
Claude Opus 5 launches — near-Fable-5 performance at half the price
Anthropic released Claude Opus 5 — a direct upgrade to Opus 4.8 at the same price. It's now the default on Claude Max and the strongest model on Claude Pro, with big gains in coding, knowledge work, and double-checking its own answers. It nearly matches the top-tier Fable 5 on most tests — so most serious work can now run on Opus 5 for half the cost. Builders: exact per-token API rates live in the Claude card.
Manus gets Plan Mode — and is now part of Meta
Manus launched Plan Mode: type
/ and select Plan (or tap + on mobile), and Manus writes a full Markdown plan — goals, steps, constraints — before it builds anything. You edit and approve, then it executes. Nothing ships until you confirm, and you can switch it on mid-task to scope the next phase. Free for all users on web and mobile. Separately, Manus confirmed it has joined Meta (after a regulator delay) and continues operating its existing products.Gemini 3.6 Flash & 3.5 Flash-Lite ship — faster and cheaper
Google released Gemini 3.6 Flash — cheaper to run than its predecessor and using ~17% fewer tokens per task — plus Gemini 3.5 Flash-Lite, the fastest model in the series for quick jobs. Both are live in the Gemini app, Google AI Studio, and Search. Still cooking: Gemini 3.5 Pro (partner testing) and Gemini 4 (pre-training). Exact API rates: see the Gemini card. A specialized Flash Cyber model stays restricted to governments/partners.
Earlier in Julykept for the month
NotebookLM is now "Gemini Notebook" — and every notebook gets a cloud computer
Google renamed NotebookLM → Gemini Notebook (same product, all notebooks preserved). It's now woven deeper into the Gemini app and is coming to AI Mode in Search. The big upgrade: every notebook now has a secure cloud computer that writes and runs code natively — so it can do real data analysis, charts, spreadsheets, and slide decks grounded in your own sources. Cloud computer is live for Google AI Ultra & eligible Workspace plans now, rolling out to Pro users over the coming weeks.
GPT-5.6 family (Sol, Terra, Luna) launches for all paid users
OpenAI launched three new models: Sol (fast/efficient), Terra (balanced), and Luna (most capable). All three are now live for Plus, Pro, Business, and Enterprise users — no plan change needed. Builders: Sol is available through the OpenAI API (rates in the OpenAI card), and real-time voice pricing was cut 30% back on July 1 — relevant if you build voice or live-conversation features.
Grok 4.5 is now publicly available — Opus-class for coding and agentic work
xAI's Grok 4.5 launched publicly on July 8, available in Grok Build, Cursor, and the xAI API console (not yet available in the EU). Builder API pricing is $2/$6 per million tokens — competitive with Claude Sonnet 5 (and a rounding error at everyday scale). xAI positions it as "Opus-class" for coding and agentic tasks. Bonus: 21 multilingual Grok Voice voices with voice cloning support also went live this month.
Claude Sonnet 5 launches — most agentic Sonnet yet, at introductory pricing
Anthropic launched Claude Sonnet 5 at a value price — their most capable Sonnet yet, designed for autonomous planning, multi-tool use, and production workflows. 🆕 Update (Aug 10, 2026): that price is now permanent — the planned September 1 increase is canceled. Exact per-token rates: see the Claude card. Note: fast mode was removed from Claude Opus 4.6 (June 29) and deprecated for Opus 4.7 on July 24.
Grok 4.1 is live for everyone — and Grok Imagine is now complete
Grok 4.1 is now rolling out to all users on grok.com, X, and the iOS/Android apps — selected automatically in Auto mode. xAI also announced on July 5 that Grok Imagine (image and video generation via the Aurora model) is complete. Bonus: the Speech-to-Text and Text-to-Speech APIs are now generally available, and a no-code Voice Agent Builder beta launched for building production voice agents in under 2 minutes.
NotebookLM and Gemini now sync — start a notebook anywhere, continue in both
Google introduced "Notebooks" — a shared knowledge base across Gemini and NotebookLM. Sources added in the Gemini app appear automatically in NotebookLM (and vice versa), so you get Video Overviews, Infographics, and AI research tools on anything you start in Gemini. Rolling out now for Google AI Ultra, Pro, and Plus subscribers on web. Also new this month: Cinematic Video Overviews, 10 new infographic styles, artifact creation in chat, EPUB uploads, and PPTX export.
Industry & Platform News
Broader shifts affecting your strategy — beyond the tools in this guide.
›
Industry & Platform News
Broader shifts affecting your strategy — beyond the tools in this guide.
Claude and ChatGPT can now connect directly to your Meta Ads account
Meta opened its ad ecosystem to third-party AI tools via MCP connectors. You can now give Claude or ChatGPT access to your Ads Manager data, pull campaign insights, and draft strategy — without leaving the chat interface. Also: the Meta AI Business Assistant is now in open beta for every advertiser and agency worldwide. Advantage+ campaigns are reporting 22% higher ROAS on average, and 4 million advertisers are now using Meta's generative AI tools.
ChatGPT is now an ad platform — open to any US business, no minimum spend
OpenAI opened self-serve advertising to all US businesses. ChatGPT processes 2.5 billion daily prompts and hit $100M in annualized ad revenue in its first 6 weeks. For coaches and consultants, this is a new top-of-funnel channel worth monitoring — the audience skews toward early AI adopters actively seeking tools and expertise. OpenAI is also building tools to let advertisers generate, localize, and optimize ad creative with AI.
Ranking #1 on Google no longer guarantees an AI Overview citation
In mid-2025, 76% of Google AI Overview citations went to top-10 ranked pages. By early 2026, that dropped to 38%. Traditional SEO and AI search visibility have officially decoupled — they now require different strategies. If your content marketing relies on organic Google traffic, this needs attention. Google AI Mode also crossed 1 billion monthly active users, and AI Overviews are at 2.5 billion — search behavior is shifting fast.
EU AI Act enters full enforcement next month — relevant if you serve EU audiences
The EU AI Act's main enforcement phase begins August 2, 2026. Tools like Claude, ChatGPT, and Gemini are already required to be transparent about AI-generated content for EU users. Meta is the notable holdout — it refused to sign the GPAI Code of Practice that Anthropic, Google, Microsoft, and OpenAI all signed. No fines have been issued yet. If you have clients or audiences in the EU, this is worth a quick check with your legal or compliance contacts.
Word, Excel & PowerPoint AI features now require a paid Copilot license
Microsoft tightened its Copilot licensing: AI features inside Word, Excel, PowerPoint, and OneNote now require a Microsoft 365 Copilot add-on (paid). If your team has been using these features under a standard M365 plan, verify your license status before August — you may hit a paywall. Free Copilot Chat remains available via the web app. On the upside, Copilot now includes GPT-5.5 Instant and Claude Opus 4.7 for those with the paid tier.
The AI landscape changes every week. Your strategy doesn't have to.
Be Known helps coaches, consultants, and expert-driven businesses cut through the noise — building AI-native marketing systems, high-converting funnels, and paid ad strategies that grow with the technology, not against it.
Which Tier Are You?
Find your level, then click any tool to learn more.
🌱
Tier 1
Just Getting
Started
Started
You're exploring AI for the first time, or you use it occasionally for simple tasks. You want something that just works — no setup, no tech knowledge required.
💬Claude.ai Chat
🔍Perplexity
💬ChatGPT
✨Gemini
⚡Grok
⚡
Tier 2
Daily User &
Experimenter
Experimenter
You use AI regularly and want to go deeper — running research workflows, automating tasks, testing prompts, or building things without writing code.
🌐Perplexity Pro
🧪Google AI Studio
📓NotebookLM
🤖Manus AI
🏗️Lovable
🖥️Claude Cowork
🚀
Tier 3
Power User &
Builder
Builder
You build systems, write prompts like a developer thinks in code, automate complex workflows, and want full control over model behavior and integrations.
💻Claude Code
🌐Perplexity Computer Use
🔧OpenAI API / GPT-4o
🔧xAI API / Grok
⚙️Replit
⚙️Base44
🔗Zapier AI
🔗Make AI
Full Tool Comparison
Click any row for full details on that tool.
Tokens 101: AI tools charge by the token — roughly ¾ of a word, and a million tokens is about 1,500 pages of text. The per-token rates you see below work out to fractions of a cent per reply, so they only really matter if you're building your own app with an API. For everyday use, the monthly plan price is what counts. Full glossary →
| Tool | Best For | Key Strengths | Limitations | Skill Level | Cost |
|---|---|---|---|---|---|
| 🟢 Anthropic · Claude Family | |||||
| Claude.ai Chatclaude.ai | Writing, strategy, copy editing, long documents | Best reasoning1M context (Opus 5)Nuanced writing | No internet by default; can't act on your computer. 🆕 Claude Opus 5 now live (July 24) — same price as Opus 4.8, default on Claude Max. Sonnet 5's value price is now permanent (no Sept increase). | 🌱 Beginner | Free / Pro $20/mo |
| Claude CodeTerminal / CLI | Writing, debugging & refactoring code; developer workflows | Reads codebaseRuns commandsEdits files | Requires developer setup; not for non-coders. ✅ Agent SDK billing change paused — usage still draws from your subscription. Claude Sonnet 5 now available via API (July 2026). | 🚀 Power User | API usage-based |
| Claude CoworkDesktop Agent | Letting Claude operate your computer on your behalf | Sees your screenTakes actionsBackground agent | Memory & CPU intensive; system must stay open | ⚡ Intermediate | Claude Pro / Teams |
| 🔵 OpenAI · ChatGPT Family | |||||
| ChatGPTchat.openai.com | General chat, brainstorming, image generation, code help | Huge ecosystemGPTs / pluginsImage gen | Quality varies; hallucinations; Pro needed for best models. GPT-5.2 Thinking retired June 5 2026. 🆕 GPT-5.6 family (Sol, Terra, Luna) launched July 9. Aug 6: Free/Go now get GPT-5.6 Luna + unlimited text chats + Think button; Plus/Pro get an updated Sol with a thinking slider. Aug 10: ChatGPT Business Premium seats ($125/seat, 5× usage). | 🌱 Beginner | Free / Plus $20/mo / Pro $100–200/mo |
| OpenAI API / GPT-5platform.openai.com | Building apps, custom chatbots, real-time voice integrations | Voice modeVisionAPI ecosystem | Requires development knowledge; cost adds up at scale. Note: GPT-4o retired April 2026. | 🚀 Power User | Usage-based API |
| ⚫ xAI · Grok Family | |||||
| Grokx.ai / X (Twitter) | Real-time chat with live X/Twitter data, unfiltered Q&A, current events, image generation | Real-time X dataLess restrictedImage gen (Aurora) | Free tier still limited; SuperGrok required for full features. 🆕 Grok 4.6 (Aug 12) is xAI's new flagship — sharp with long, multi-step tasks. Same $30/mo plan pricing; API rates for builders are in the Grok card. | 🌱 Beginner | Free / SuperGrok $30/mo |
| xAI API / Grokconsole.x.ai | Building apps with Grok models; accessing real-time X data programmatically; large context tasks | 1M contextReal-time X feedFunction calling | Smaller developer ecosystem; less mature tooling than OpenAI or Anthropic APIs | 🚀 Power User | Usage-based API pricing |
| 🟣 Perplexity Family | |||||
| Perplexityperplexity.ai | Real-time research, cited answers, fact-checking | Live web searchCited sourcesFast research | Less suited for creative writing or long-form generation | 🌱 Beginner | Free / Pro $20/mo / Max $200/mo |
| Perplexity Computer UsePro feature | Autonomous web browsing, multi-step research tasks | Browses the webMulti-step tasksResearch agentBrain memory (July 2026) | Still maturing; limited to browser actions. 🆕 Personal Computer expanded to Windows (July 28); Brain self-improving memory added early July. Aug 6: shared Projects workspaces and expanded Model Council (2–8 models, now on Pro); GPT-5.6 Terra/Luna now live in Computer. | 🚀 Power User | Perplexity Max ($200/mo+) |
| 🔴 Google AI Family | |||||
| Geminigemini.google.com | Everyday AI chat, Google Workspace integration | Google integration1M token contextMultimodal3.7 Flash (Aug 2026) | Less nuanced reasoning than Claude; ecosystem lock-in. 🆕 Gemini 3.7 Flash shipped Aug 13 — faster and about half the API cost of the previous version through Dec 31; 3.5 Pro still in testing. API rates: see the Gemini card. | 🌱 Beginner | Free / AI Pro $19.99/mo |
| Google AI Studioaistudio.google.com | Testing prompts for free, prototyping, saving workflows | Free sandbox1M tokensPrompt testing | Developer-oriented UI; Gemini models only | ⚡ Intermediate | Free (generous limits) |
| Gemini Notebooknotebooklm.google.com | Uploading documents and asking questions about them | Document Q&AAudio & Video OverviewsSource-groundedCharts & slide exportsNow Gemini Notebook (July 2026) | Limited to your uploaded sources; not general-purpose. 🆕 Renamed Gemini Notebook (July 16); cloud computer for code is live for Ultra/Workspace, rolling out to Pro. | ⚡ Intermediate | Free |
| 🟠 AI App & Web Builders | |||||
| Lovablelovable.dev | Building full web apps from a prompt — no coding required | Full apps from textSupabase integrationFast iteration | Less control than coding from scratch; best for MVPs. Credit-based — complex features cost more than simple edits. | ⚡ Intermediate | Free / Pro $25/mo |
| Replitreplit.com | Running, editing, deploying code in the browser | Cloud IDEDeploy instantlyAI autocompleteMCP + security scan (Aug 2026) | Effort-based credits (since Feb 2026) — complex tasks burn fast. Slower than local dev for large projects. 🆕 Aug 14: MCP server (ChatGPT/Claude/Slack), workspace regions, Clerk Auth/SSO, beta Admin API; Aug 7: auto Semgrep security scan. | 🚀 Power User | Free / From $25/mo |
| Base44base44.com | AI-powered business app builder with built-in database & auth | Built-in DBNo backend neededBusiness appsBranches (Aug 2026) | Acquired by Wix (2025); credits don't roll over; smaller community than Lovable or Replit. 🆕 Aug 9: Branches — build on a separate version and merge to main when ready (up to 5 parallel builds), on Builder plan+. | ⚡ Intermediate | Free / From $16/mo |
| 🤖 Autonomous AI Agents | |||||
| Manus AImanus.im | Fully autonomous research and multi-step task execution | Autonomous agentMulti-step tasksMCP integrationsPlan Mode (July 2026) | Slower; results need review; no conditional logic or integration triggers like Zapier. 🆕 Plan Mode live July 22; Manus now part of Meta. | ⚡ Intermediate | Free / From $20/mo |
| ⚙️ AI-Powered Automation | |||||
| Zapier AIzapier.com | Connecting apps and using AI to make decisions in automations | 6,000+ integrationsAI stepsNo-code | AI steps priced in three tiers (Standard/Advanced/Premium) — higher tiers cost more per AI step, so keep an eye on usage costs | ⚡ Intermediate | Free tier / From $19.99/mo |
| Makemake.com | Visual automation with more control and lower cost than Zapier | Visual flow builderCheaper at scaleMore flexibility | Steeper learning curve; fewer native integrations | ⚡ Intermediate | Free / From $9/mo |
Use-Case Quadrant
Click any dot to learn more about that tool.
X-axis: Task Type · Y-axis: How much the AI acts on its own vs. you staying in control
Dark green = Most autonomous / technical
Mid green = Intermediate
Bright green = Accessible / assisted
Glossary: Key Terms Explained
Plain-English definitions for the jargon you'll encounter.
LLM
Large Language Model. The AI "brain" behind most chat tools. Examples: GPT-4, Claude 3, Gemini. They predict and generate text based on your input.
Model
The specific version of an AI. Claude Sonnet vs. Opus, GPT-4o vs. GPT-3.5. Bigger models = more capable but slower and costlier.
Prompt
The instruction or question you type to the AI. Better prompts = better results. Think of it as how you talk to the tool.
API
Application Programming Interface. A way for developers to connect AI models directly into their own apps or workflows. Requires coding or automation tools.
Token
How AI measures text. Roughly 1 token = 1 word (or part of a word). API costs are often charged per token used. More tokens = more cost.
Context Window
How much text the AI can "see" and remember at once in one conversation. Claude and Gemini lead here with 100K–1M token windows.
MCP
Model Context Protocol. A standard that lets AI connect to external tools and services — like your Google Drive, Slack, or ad accounts — so it can read and act on real data.
Agent / Agentic AI
An AI that doesn't just answer questions — it takes actions. Agents can browse the web, write files, click buttons, or complete multi-step tasks on your behalf.
Computer Use
A capability that lets an AI see and control a computer screen — clicking, typing, navigating — as if it were a human at the keyboard.
Skill / System Prompt
Hidden instructions that tell an AI how to behave before you ever type a message. Used to give AI a role, set rules, or load context automatically.
Temperature
A setting that controls how "creative" vs. "precise" the AI is. Low temperature = consistent, factual. High temperature = more surprising, varied output.
Hallucination
When an AI makes up something that sounds plausible but isn't true. All LLMs do this. Always verify important facts, especially with ChatGPT and older models.
RAG
Retrieval-Augmented Generation. A technique where AI pulls from your own documents or database before answering — reducing hallucinations. NotebookLM uses this.
Fine-tuning
Training an existing AI model on your own data to make it better at your specific tasks. Requires technical setup. Usually done via API.
Multimodal
An AI that can work with more than just text — images, audio, video, documents. ChatGPT, Claude, and Gemini are all multimodal.