ChatGPT

From PRIMO.ai
Jump to navigation Jump to search

YouTube ... Quora ...Google search ...Google News ...Bing News


Generates human-like text, based on a family of “large language models (LLM)” — algorithms that can recognize, predict, and generate text based on patterns they identify in datasets containing hundreds of millions of words; ChatGPT performs a wide range of natural language processing (NLP) tasks; chatbots, automated writing, language translation, text summarization and generate computer programs. OpenAI states ChatGPT is a significant iterative step in the direction of providing a safe AI model for everyone. ChatGPT interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer follow-up questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests. ChatGPT is a sibling model to InstructGPT, which is trained to follow an instruction in a prompt and provide a detailed response.


ChatGPT will not take your jobs—someone who knows how to use it will. - Jaspreet Bindra


What is ChatGPT?

Load video
YouTube
Reinforcement Learning from Human Feedback: From Zero to ChatGPT
In this talk, we will cover the basics of Reinforcement Learning (RL) from Human Feedback (RLHF) and how this technology is being used to enable state-of-the-art ML tools like ChatGPT. Most of the talk will be an overview of the interconnected ML models and cover the basics of Natural Language Processing (NLP) and Reinforcement Learning (RL) that one needs to understand how RLHF is used on large language models. It will conclude with open question in RLHF.

Nathan Lambert is a Research Scientist at Hugging Face. He received his PhD from the University of California, Berkeley working at the intersection of machine learning and robotics. He was advised by Professor Kristofer Pister in the Berkeley Autonomous Microsystems Lab and Roberto Calandra at Meta AI Research. He was lucky to intern at Facebook AI and DeepMind during his Ph.D. Nathan was was awarded the UC Berkeley EECS Demetri Angelakos Memorial Achievement Award for Altruism for his efforts to better community norms.

Load video
YouTube
But How Does ChatGPT Actually Work?
You’ll learn how ChatGPT works and this will provide many benefits, such as helping you to use the model more effectively, evaluate its outputs more critically, and staying informed about the latest developments in the field so you are better prepared to take advantage of new opportunities. ChatGPT is a type of Natural Language Processing (NLP) known as a Generative Pre-trained Transformer (GPT) developed by OpenAI. These are the two big terms we will focus on in this video. On top of that you will also get a base understanding of common Machine Learning techniques like Supervised Learning, and Reinforcement Learning (RL), which were used to make ChatGPT as good as it is.

Load video
YouTube
How ChatGPT is Trained
This short tutorial explains the training objectives used to develop ChatGPT, the new chatbot language model from OpenAI.

Timestamps:

  • 0:00 - Non-intro
  • 0:24 - Training overview
  • 1:33 - Generative pretraining (the raw language model)
  • 4:18 - The alignment problem
  • 6:26 - Supervised fine-tuning
  • 7:19 - Limitations of supervision: distributional shift
  • 8:50 - Reward learning based on preferences
  • 10:39 - Reinforcement learning from human feedback
  • 13:02 - Room for improvement

ChatGPT: https://openai.com/blog/chatgpt

Special thanks to Elmira Amirloo for feedback on this video.

Links: YouTube: https://www.youtube.com/ariseffai Twitter: https://twitter.com/ari_seff Homepage: https://www.ariseff.com

If you'd like to help support the channel (completely optional), you can donate a cup of coffee via the following: Venmo: https://venmo.com/ariseff PayPal: https://www.paypal.me/ariseff

Load video
YouTube
ChatGPT in Architecture EXPLAINED
If you're looking to revolutionize your architecture workflow, then you need to know about ChatGPT. Developed by OpenAI, this advanced AI model with an impressive 175 billion parameters is making waves in the industry, and it's not hard to see why. In just five days, it got over 1 million users, outpacing Facebook by ten months. But what is ChatGPT and how can it benefit architects? In this video, we dive deep into the technology, its uses, and how it can help you save time and create innovative designs.

Discover the potential of ChatGPT for answering questions, writing essays, generating content ideas, writing scripts and titles for YouTube videos, and much more. The possibilities are endless, and as we explore this fascinating technology, you'll learn how it can help you generate conceptual design ideas faster and refine them before moving to the next phase of your project. We'll also explore the connections ChatGPT has with Rhino and Grasshopper, the impact on creative industries, and the current state of the technology.

Load video
YouTube
ChatGPT Architecture Explained: Step-by-Step Guide
About Presenter: Sandeep Giri. Past - Amazon, InMobi, tBits Global, D.E.Shaw. For last 16 years, Sandeep has been building products and churning large amounts of data for various product companies. He has an all-around experience in software development and big data analysis. Apart from digging data and technologies, Sandeep enjoys conducting interviews and explaining difficult concepts in simple ways. About CloudxLab Email us at: reachus@cloudxlab.com Official Website link: https://cloudxlab.com/

Load video
YouTube
ChatGPT Tutorial for Developers - 38 Ways to 10x Your Productivity
Learn how to use ChatGPT to 10x your productivity! 38 examples using Python, JavaScript, HTML, CSS, React, SQL and more!

Load video
YouTube
Load video
YouTube
Load video
YouTube
Load video
YouTube

2024-2025 Milestones

The 2024-2025 period marked a significant shift in ChatGPT's architecture and user interface:

  • Transition to Reasoning Models: OpenAI introduced the o1 series and subsequent o3 models, which are specifically trained to "think" before they speak using chain-of-thought reasoning. These models excel at complex tasks in science, coding, and mathematics by breaking down problems into logical steps, distinguishing them from standard chat models that prioritize immediate response generation.
  • Canvas Interface: OpenAI launched Canvas, a dedicated workspace for collaborative writing and coding projects. Unlike the standard chat interface, Canvas allows users to edit documents and code snippets directly in a side-by-side window, enabling more precise control, iterative refinement, and context-aware collaboration.
  • Advanced Multimodal Integration: The integration of GPT-4o unified text, audio, and vision, allowing for real-time, low-latency interactions that feel more natural and responsive.

Multimodal Capabilities

ChatGPT has evolved from a text-only interface to a fully multimodal system, supporting end-to-end speech, real-time vision processing, and advanced image generation.

  • Vision: Users can upload images, screenshots, or documents. ChatGPT uses vision models to analyze, describe, and solve problems based on visual input, now with enhanced accuracy in the desktop and mobile applications.
  • Voice Mode: Recent updates, particularly with GPT-4o, introduced real-time, low-latency voice interaction. The system can detect emotional cues, adjust its tone, and handle interruptions, providing a fluid conversational experience.
  • Real-time Interaction: The model's ability to process audio and visual data in real-time allows for use cases like live language translation, visual assistance for the visually impaired, and interactive tutoring.

Agents

YouTube ... Quora ...Google search ...Google News ...Bing News

An AI agent is software that pursues a goal by planning steps, using tools, and acting on a person's behalf, rather than answering one question at a time. Unlike traditional robotic process automation, which follows fixed scripts, an agent built on a large language model decides what to do next based on what it finds, so it can open a browser, run code, read files and connected apps, produce finished deliverables, and keep working while the user is away. OpenAI spent 2025 and 2026 turning ChatGPT from a conversation partner into a place where work gets delegated, and its desktop app now offers three modes: Chat for answers and drafts, Work for multi-step projects that end in a reviewable file, and Codex for software engineering. A fourth product, dots, sits above all three as an always-on agent that can hand tasks to Work or Codex and follow up on them over days or weeks.

The two consumer-facing agents covered below solve different problems. ChatGPT Work, launched on July 9, 2026, is a delegate for a defined task: you describe an outcome, it gathers context, breaks the job into steps, and returns finished sheets, slides, documents, or web apps. Dots, announced at OpenAI DevDay on September 29, 2026, are persistent agents with their own cloud computer and browser that take ongoing responsibility for an area of work and check in only when a decision needs a human. Both rely on the same plugin connections, approval checks, and GPT-5.6 and GPT-6 family models, so what a beginner learns about briefing one carries over to the other.

OpenAI's Agent Efforts

OpenAI's agent work has advanced along three tracks that steadily converged. The first track put agents inside ChatGPT for everyday knowledge work, starting with the browser-driving Operator and the deep research tool, then merging them into ChatGPT agent and later into ChatGPT Work and dots. The second track built coding agents under the Codex name, which began as a command-line tool and a cloud service and eventually became one of the three modes of the ChatGPT desktop app. The third track gave developers and enterprises the building blocks for their own agents, from the Responses API and Agents SDK to AgentKit, the Frontier enterprise platform, and a hosted Agents API. The table below lists the major milestones.

Date Release What it added
Jan. 23, 2025 Operator A research-preview agent that used its own web browser to fill out forms, place orders, and complete other repetitive web tasks
Feb. 2, 2025 Deep research Multi-step web research that reads many sources and returns a cited report
Mar. 11, 2025 Responses API and Agents SDK Developer building blocks for tool use, web and file search, computer use, and multi-agent orchestration
Apr. 16 and May 16, 2025 Codex CLI and Codex cloud An open-source terminal coding agent, followed by a cloud agent that writes features, fixes bugs, and proposes changes for review
July 17, 2025 ChatGPT agent One "agent mode" that combined Operator's browser actions with deep research, a terminal, and connectors to apps
Oct. 6, 2025 AgentKit and Apps SDK (DevDay 2025) Agent Builder (a visual canvas for multi-agent workflows), ChatKit, a Connector Registry, evals, guardrails, and third-party apps that run inside ChatGPT
Oct. 21, 2025 ChatGPT Atlas A web browser with ChatGPT and agent mode built in
Feb. 5, 2026 Frontier An enterprise platform to build, deploy, govern, and evaluate "AI coworkers," including agents from other vendors
July 9, 2026 ChatGPT Work and GPT-5.6 The successor to ChatGPT agent, plus Scheduled Tasks, Sites, and one desktop app combining Chat, Work, and Codex; OpenAI also announced that Atlas would be retired
Sept. 2026 GPT-6 family and Agents API GPT-6 Astra (Sept. 3), GPT-6 Sol and GPT-6 Luna (Sept. 22), and a public beta of a hosted Agents API built on the same harness as Codex
Sept. 29, 2026 DevDay 2026 Dots, ChatGPT Space, GPT-6.1 Sol, computer use and multi-agent support in the Agents API, and a preview of the Decisions API

Several lessons from this history shape how today's products behave. Operator and ChatGPT agent showed that a model controlling a browser could finish real errands, but they also exposed the risk that a malicious web page could hijack an agent through a prompt injection attack, which is why OpenAI required confirmation before consequential actions and later added automated "auto-review" of risky steps. OpenAI trained ChatGPT agent with reinforcement learning to choose the right tool for each step, an approach that carries into Work and dots. The merger of Codex into the ChatGPT desktop app in July 2026 reflected another lesson: more than 1 million of Codex's roughly 5 million weekly users were already using it for work that had nothing to do with software, so OpenAI put the same agent harness behind a friendlier, non-technical front door.

Recent Developments

The pace accelerated in the fall of 2026. GPT-6 Astra arrived on September 3 as OpenAI's flagship model for computer use, software engineering, and long-horizon agent work, followed by the cheaper GPT-6 Sol and GPT-6 Luna on September 22 and an upgraded GPT-6.1 Sol at DevDay a week later. At DevDay, OpenAI also expanded its hosted Agents API with computer use, multi-agent orchestration, tool search, and context compaction, previewed a Decisions API that limits a model to a finite set of predefined answers for classification and routing, and made the same agent features available inside AWS through Amazon Bedrock Managed Agents. The company introduced ChatGPT Space as well, a shared project area where teammates, ChatGPT, and a dot work from the same knowledge, including Pages (documents that people and agents write together) and collaborative slides.

Smaller updates have filled in gaps for both agents. ChatGPT Work gained the ability to complete tasks on some websites that require signing in (August 2026), event-triggered Scheduled Tasks that respond to changes in Slack, Gmail, and GitHub, a Data plugin for building dashboards from connected business data, and, beginning October 26, 2026, the GPT-6.1 Terra model in Work and Codex. Dots gained texting over iMessage or RCS on October 28, 2026, with the texting channel sharing the same memory and plugins as the dot in the ChatGPT app. Because these products change almost weekly, treat the plan and model details in this section as a snapshot from late October 2026 and confirm them against OpenAI's release notes.

ChatGPT Work vs. Dots: Key Differences

The simplest way to tell the two apart is to ask whether the work has a finish line. OpenAI's documentation recommends ChatGPT Work for delegating a task, such as research, analysis, or creating a file, either now or on a schedule, and recommends dots for tracking an ongoing responsibility, deciding next steps as circumstances change, and following up between conversations. A Work task is like handing a well-defined assignment to a skilled contractor: you brief it, it works for minutes or hours, and it hands back a deliverable for review. A dot is closer to the "chief of staff" that Sam Altman described at DevDay, a standing assistant that learns your preferences, watches the sources you connect, and decides when something deserves your attention. The two are complementary rather than competing, since a dot can delegate pieces of its responsibility to Work or Codex and track the results.

Feature ChatGPT Work Dots
Core idea Delegate a task with a clear, reviewable outcome Hand off an ongoing responsibility that continues between conversations
Launched July 9, 2026 (successor to ChatGPT agent) September 29, 2026 (DevDay 2026)
Model Selectable; launched on GPT-5.6 and now offers GPT-6 Astra, GPT-6.1 Sol, and Luna, with GPT-6.1 Terra rolling out GPT-6 Astra
Where it runs On your computer through the desktop app (approved local files, apps, and browser) or in an OpenAI cloud container Its own cloud computer and browser, which keep working when your computer is off; optional access to one connected personal computer
How work starts You start each job, or create a Scheduled Task that runs once, on a schedule, on an event, or to monitor for changes You assign responsibilities; it also researches connected apps proactively (read-only) and suggests next steps
Where you talk to it ChatGPT desktop app, web, and mobile ChatGPT (text or voice, on desktop, web, and mobile), Slack, Microsoft Teams, and text messages
Typical output Finished spreadsheets, slide decks, documents, reports, and Sites (shareable web apps) Briefings, drafts, updated documents, follow-ups, and tasks it delegates to Work, Codex, or subagents
Memory Uses the files, plugins, and project context you give the task Starts from your ChatGPT memory and keeps its own notes on preferences, decisions, and ongoing work
Controls Choose what it can access, follow progress, and approve important actions; auto-review checks risky actions Built-in approval rules, optional Custom Rules, auto-review, safety monitoring, Activity and Scheduled views, pause, and delete
Who can use it Desktop app: every plan, including Free; web and mobile: paid plans other than Free and Go Pro (ages 18+, outside the EEA, UK, and Switzerland), Business Premium, and Enterprise, Edu, or Healthcare workspaces when an admin enables it
Usage Metered like Codex; longer and more complex tasks consume more of a plan's allowance or workspace credits Conversations with a dot don't count toward ChatGPT limits; deeper background work draws on a plan allowance; tasks it delegates count toward Work or Codex limits

A practical rule of thumb for beginners follows from the table. If you can describe the finished file you want in a paragraph, start with ChatGPT Work, which is available to more people (including free users of the desktop app) and gives you a concrete artifact to judge. If what you want is "keep an eye on this and tell me when it matters," and you have an eligible plan, a dot is the better fit, provided you are comfortable granting a persistent agent access to your calendar, inbox, or chat channels. Many people will end up using both: Work for projects and deliverables, and a dot as the coordinator that remembers context, schedules follow-ups, and hands well-scoped tasks to Work.

ChatGPT Work

ChatGPT Work is OpenAI's agent for longer, multi-step knowledge work inside ChatGPT. OpenAI introduced it on July 9, 2026, as an agent that can act across a user's apps and files, stay with a project for hours, and turn a goal into finished spreadsheets, presentations, documents, and web apps. It replaced the earlier ChatGPT agent and runs on Codex technology, the agent harness OpenAI first built for software engineering, paired with a model chosen from the model selector. Work breaks a goal into smaller steps, completes them independently, and carries context from one step to the next, so a single request can move from customer research to a campaign brief to market-specific marketing assets without the user restating the background each time. OpenAI says Work cut its own month-end close and forecasting from days to hours, and Zapier reported using it to review thousands of leads and build an executive dashboard that surfaced seven figures in potential sales.


ChatGPT Work turns the chat window into a delegated project workspace: it gathers context across your connected apps and local files, runs multi-step analysis, and hands back finished documents rather than rough drafts.


Work gathers context through plugins, which connect ChatGPT to services such as Slack, Microsoft Teams, Google Drive, SharePoint, email, calendars, CRM systems, project trackers, and internal tools. ChatGPT picks relevant plugins automatically, or you can point it at one by typing "@" followed by the plugin's name, and plugin permissions are shared across Chat, Work, Codex, and dots. In the desktop app, Work can also use local files and applications, a built-in browser for research, and a background computer-use capability that clicks, types, and moves files on the user's behalf. On web and mobile, Work runs in an OpenAI cloud container, and since August 2026 its browser can complete tasks on some websites that require signing in, so a user can hand off a task and step away. With Canvas retired from ChatGPT in 2026, Work has also become the main place for long-form document and analysis work.

Two companion features extend what Work can do. Scheduled Tasks let ChatGPT run an action once, repeat it on a schedule, react when an event occurs, or monitor something for changes, such as reviewing a Slack channel each week to refresh a meeting agenda or updating a deck when feedback arrives by email; free accounts can keep up to three active scheduled tasks that run at most once a day, while paid plans such as Plus and Pro get more and can trigger tasks from changes in Slack, Gmail, and GitHub. Sites, in public beta, turns work into an interactive website or lightweight app, such as a live dashboard, project tracker, or prototype, that can be shared with a team or publicly by URL and updated as the underlying information changes. Sites is not available on the Free or Go plans, and public publishing initially excluded the EEA, Switzerland, and the United Kingdom.

Usage works differently from ordinary chat. Work is metered like Codex: longer and more complex tasks, more capable models, faster speed settings, and heavy tool use all consume more of a plan's included usage, and eligible business agreements draw on a pool of workspace credits that administrators can cap with defaults, group limits, individual overrides, and alerts. In the ChatGPT desktop app for Mac and Windows, Chat, Work, and Codex are available on every plan, including Free, while on the web and mobile apps Work is limited to paid plans other than Free and Go. Enterprise and Edu workspaces received a two-week preview period in which Work was off by default, and their administrators control which people, plugins, websites, and actions Work may use. For model choice, OpenAI suggests GPT-6 Astra for demanding work that needs careful reasoning or a polished final file, GPT-6.1 Sol for complex work at lower cost, and Luna for simpler jobs.

Work keeps the user in the loop by design. You decide what it may access, can watch its progress, answer its questions, change direction mid-task, and approve important actions, while an auto-review step uses advanced models to check consequential actions involving connected tools before they happen, aiming to stop unauthorized sharing of sensitive information. By default the agent asks before visiting a new website, and workspace administrators can restrict apps to read-only access or approved actions. The system still makes mistakes, so OpenAI's guidance is to ask for a draft first, have Work flag unsupported claims or items that need human review, and check the output before sharing it or using it for decisions. Practical limits also apply: a cloud task cannot reach files on a computer that has gone offline, a task cannot switch between local and cloud execution in the middle of a turn, and enterprise cloud browsing cannot use saved passwords or local browser sessions.

Getting Started with ChatGPT Work

A beginner can get a first useful result from Work in a single short session. OpenAI's advice is that good first tasks have a clear outcome, a few source materials, and an output you can review, so it pays to start with something you know well enough to judge. The steps below use the desktop app because it is available on every plan, but the web and mobile apps work the same way on paid plans.

  1. Install and sign in. Download the ChatGPT desktop app for Mac or Windows from openai.com/chatgpt/desktop and sign in. The older desktop app has been renamed ChatGPT Classic, so make sure you open the new one.
  2. Switch to Work. Select ChatGPT (not Codex) from the menu, then choose Work from the Chat and Work toggle. On the web or mobile, select Work at the top of a new chat. To reuse a project's files, open the project and choose Work there.
  3. Pick a model and check your allowance. Use the model selector; Luna or GPT-6.1 Sol is a sensible choice for a first task, and Free and Go users can use GPT-6 Luna in the desktop app. Eligible plans show remaining Work usage in the sidebar.
  4. Connect a plugin (optional). Select Customize, then Plugins, in the left navigation and connect one service you use, such as Google Drive or Gmail. You can skip this step for the working example below.
  5. Write a brief, not a question. State the outcome, the sources it may use, constraints, what "good" looks like, and when it must stop for your review. This is ordinary prompt engineering applied to delegation.
  6. Follow along and approve. Watch the progress view, answer clarifying questions, redirect it if it drifts, and approve any important actions.
  7. Review the file. Open the spreadsheet, deck, or document it produces, spot-check facts and numbers, and ask for revisions in the same conversation.

Working Example: A Laptop Purchase Comparison Spreadsheet

This example needs no plugins and no outside data, so it works on any plan that includes Work, including the free desktop app. It mirrors one of OpenAI's recommended first tasks, a comparison spreadsheet that scores options and recommends a choice, but it supplies sample data in the prompt so every reader gets a result that can be checked by hand. Copy the brief below into a new Work conversation. 

Goal: Help me choose a laptop for our 5-person design team.

Sources: Use only the data below. Do not browse the web.

Options (price in USD, battery life in hours, weight in kg, RAM in GB, warranty in years):
Option A: 1899, 18, 1.6, 32, 3
Option B: 1499, 12, 1.9, 16, 1
Option C: 2199, 22, 1.4, 24, 2

Deliverable: an Excel workbook with three tabs.
1. "Data": the table above with clear column headers.
2. "Scoring": score each option from 1 to 5 on cost, battery, weight, RAM,
   and warranty. Put the weights (cost 30%, battery 20%, weight 15%,
   RAM 25%, warranty 10%) in their own cells and calculate weighted totals
   with formulas, so I can change a weight and see the ranking update.
3. "Summary": the ranked shortlist, the total cost of buying 5 laptops of
   each option, a one-paragraph recommendation, and a list of risks or
   missing information I should check before buying.

Constraints: State any assumption you make about how scores are assigned.
Stop and show me the draft workbook before doing anything else.

Work will typically plan its approach, build the workbook, and present it as a file you can open, often explaining the scoring method it chose. Check the result the way you would check a colleague's work. The total-cost figures are easy to verify by hand (Option A: $9,495; Option B: $7,495; Option C: $10,995), and changing the cost weight from 30% to 50% should visibly change the weighted totals if Work used formulas as asked. If a number is wrong or a score looks arbitrary, say so in the same conversation, for example "Explain how you scored battery life and switch to a linear scale," and Work will revise the file.

Once the first result looks right, follow-up requests show how Work chains deliverables while keeping context. Asking "Turn the Summary tab into a three-slide deck for my manager, using the numbers from the workbook" produces a presentation from the same analysis, and on a paid plan "Publish this comparison as a Site my team can view" turns it into a shareable page. After you connect plugins such as Slack and Google Drive, a Scheduled Task can keep the analysis current, as in the brief below. 

Every Monday at 9:00 AM Eastern for the next four weeks, review new messages
in the #laptop-purchase Slack channel and new files in the "Laptop Purchase"
Google Drive folder. Update the workbook if prices or options change, and send
me a summary of what changed here in ChatGPT. Do not post in Slack or share
the file with anyone.

Getting Started: ChatGPT Desktop Voice, Computer Use, and a Persistent Chief of Staff

A practical chief-of-staff workflow combines conversation, delegated tasks, access to applications, and a record of decisions. Voice supplies the conversational interface; Work carries out defined assignments; Computer Use operates an application's visible interface; and dots can maintain ongoing responsibilities between conversations. Calling a chat a "chief of staff" defines its role, but does not by itself enable monitoring, connect accounts, or create a recurring task. Configure those capabilities explicitly and test them with a small assignment before relying on them for daily work.

Install, connect, and test

  1. Install the desktop app. Use the download options in OpenAI's desktop app guide, sign in, and check that your account offers Work and Voice. Availability varies by plan, platform, rollout, and workspace policy. Downloading the app does not mean the models run offline.
  2. Create a workspace for the trial. Make a folder named "Chief of Staff Demo" in Documents and start a Work chat with access to that folder. Put only the documents needed for the exercise there. Choose a descriptive chat name so you can return to it.
  3. Start a live conversation. Select the Voice control in a new chat before sending the first message. Allow microphone access and choose a voice when prompted. Where supported, an existing task offers "Start voice chat"; dictation instead converts speech to message text.
  4. Test understanding. Say, "Repeat my goal in one sentence and list the information you still need." Check the visible transcript for names, dates, and quantities. Correct mistakes before the agent uses them in a file or application.
  5. Enable Computer Use. In Work or Codex, open Plugins, install or enable Computer Use, and turn on its server and skill if those switches appear. Review application access under Settings > Computer use. On macOS, grant the requested Screen Recording and Accessibility permissions; on Windows, keep the target application visible on the active desktop.
  6. Connect the services needed for the exercise. Connect Gmail and Google Calendar if available to your account, and verify the selected accounts. Prefer their structured plugins for searches and calendar reads. Use Computer Use for the Notepad exercise or another task that actually requires the visible interface.
  7. Try one harmless action. Open a blank Notepad window and say, "Use Computer Use to enter the heading Chief of Staff Demo in this blank Notepad document. Save it in the demo folder as briefing.txt, then confirm the saved path." Check the file yourself before adding more applications.

On Windows, Computer Use occupies the foreground and can move the pointer and type into the active desktop. Do not try to work in the same Windows session while it is controlling an application; pause the task first or use another device. Local desktop work also depends on the computer remaining available, whereas a dot's cloud work can continue independently of that computer. These are different execution environments, so a cloud agent should not be assumed to have access to an open local Notepad window.

Give the assistant a concrete operating brief

The following is an example instruction, not a built-in product mode. Adapt the folder, time zone, sources, and approval boundaries to your own work. Keeping a short written brief makes the workflow easier to resume and gives you something concrete against which to judge the result. 

Act as my chief of staff for this project.
Goal: maintain a reliable daily brief and an accurate list of open actions.
Sources: the Chief of Staff Demo folder, my connected Gmail account, and
my primary Google Calendar. Use America/New_York for dates and times.
Start with a one-time trial for today. Do not create recurring work yet.

You may read the named sources and create new files in the demo folder.
For this trial, do not send email, invite people, change calendar events,
delete files, purchase anything, or publish or share information.
Prepare drafts and proposals for my review.

Keep a status note containing the goal, decisions, completed work,
open actions, blockers, source links, and the next step.
Separate facts from suggestions. If information is unavailable, say so.
Before claiming completion, open each saved output and check it.
Report the output path and anything that remains incomplete.

Walkthrough: a morning brief across Notepad, Gmail, Calendar, and files

This exercise produces a local briefing and a status note without sending messages or changing anyone's schedule. Begin with the one-time trial above and keep the target Notepad window open. If Gmail or Calendar is unavailable, supply a small text export and ask the assistant to label the exercise as based on that export, rather than claiming it checked the live account.

  1. Define the time window aloud. Say, "Prepare my morning brief for today in America/New_York. First confirm today's date and which accounts you can actually access." Resolve any date or account mismatch before continuing.
  2. Read the calendar. Say, "Read today's events from my primary calendar. List start and end times, meeting titles, and preparation needs. Flag overlaps, and keep all-day events separate." Review the list against the calendar; an empty result and an inaccessible calendar must be reported differently.
  3. Read a bounded email set. Say, "Find messages received since 5 PM yesterday that need an answer or relate to today's meetings. Give me up to five important items with sender, subject, received time, source link, and a suggested next action. Do not send anything." Ask it to explain why each item needs attention and to distinguish an explicit deadline from an inferred priority.
  4. Read local priorities. Put a short file named priorities.txt in the demo folder with three genuine tasks. Say, "Read priorities.txt and combine it with the calendar and email findings. Propose my top three actions and explain any conflict." This step makes the assistant reconcile sources rather than merely produce separate summaries.
  5. Produce the brief in Notepad. Say, "Use Computer Use to update the demo Notepad document with today's date, my three priorities, today's meetings, email follow-ups, and unresolved questions. Save it as briefing.txt in the demo folder. Do not overwrite a different document." If a save or overwrite dialog appears, check that the filename and folder are the intended ones.
  6. Verify the saved result. Say, "Reopen briefing.txt from disk. Check that the last section is present, the times match the calendar, and the email links point to the messages you used. Tell me what you could not verify." Inspect the file yourself and correct errors in the same conversation.
  7. Prepare one follow-up. Say, "Draft a reply for the highest-priority email and put the draft in the brief for review. Do not send it." Review the recipient, commitments, and any quoted facts before separately authorizing a send.
  8. Record the handoff. Say, "Save a status note with completed steps, decisions, open actions, blockers, and the exact next step. Mark suggestions as suggestions and completed actions as completed only when you have evidence." End the voice conversation and return to the saved note to check that the workflow can be resumed.

A successful trial ends with readable files, traceable source items, and an explicit list of unresolved questions. It does not require the assistant to invent an urgent message or manufacture a calendar conflict. If no message needs a response, the brief should say so; if a source failed, it should identify the missing source and the resulting limitation. These checks are useful habits for any delegated workflow involving email, documents, or scheduling.

Make the responsibility persistent

For a recurring deliverable, create a scheduled Work task and inspect the saved schedule. For an ongoing responsibility that requires follow-up as circumstances change, use a dot when available, define its remit, and connect the sources it needs. A reusable chat or saved note provides continuity, but is not evidence that an autonomous monitor is running. Keep these three things distinct: saved context, a scheduled trigger, and an active ongoing assignment. 

Set up a weekday morning brief at 8:00 AM America/New_York.
Use only the accounts and folder named in our operating brief.
Show me the saved schedule, its time zone, and where results will appear.
If the schedule cannot be created, say so rather than promising to run it.
For this recurring task, prepare the brief here; do not send email or
change calendar events. Identify inaccessible sources in each brief.

For a dot, start with one narrow responsibility such as watching a project's deadlines. State which changes merit an update and where that update should appear; connecting a messaging channel alone does not establish a monitoring instruction. If local applications are necessary, connect the computer through the dot's profile and keep the computer online with ChatGPT open. Review the dot's Activity view for actual tasks, results, and requests for input, and stop or revise the assignment when it is no longer useful. 

Ongoing responsibility: track the deadlines in my project status note
and the related calendar events. Ask before changing any event.
Keep routine notes here in ChatGPT. Notify me when a deadline changes,
a conflict appears, a task finishes or fails, or you need my decision.
Stay quiet when nothing meaningful has changed.
Tell me which sources are connected and what would stop monitoring.


Videos for this workflow

Load video
YouTube

Ep 877: ChatGPT Voice is Like Jarvis: How to use the New Feature and the 7 biggest unlocks is published by Everyday AI. The latest ChatGPT Voice update acts as an advanced agent management interface, essentially providing users with a personalized version of Jarvis. By integrating the GPT Live model into the desktop version of ChatGPT Work (also known as Codeex), this feature allows for full duplex, hands-free operation. Unlike previous dictation or walkie-talkie style voice modes, this system enables the AI to listen, speak, and coordinate tasks simultaneously while actively controlling desktop applications. Key advantages of this new capability include the ability to access and utilize a user's data, connectors, and Model Context Protocol (MCP) servers, which standard voice models cannot do. This connectivity allows the AI to perform complex actions such as triaging emails, organizing files, and executing multi-step workflows. Because it can be controlled remotely via an iPhone, users can effectively manage their digital environments from anywhere, leading to significant productivity gains by reducing the distraction often associated with manual screen interaction.

Load video
YouTube

ChatGPT Voice is the most powerful AI feature of 2026. Here's how to master it features Alex Finn, The video highlights the transformative power of ChatGPT Voice, positioning it as a sophisticated orchestrator for digital productivity rather than a simple dictation tool. Unlike other voice assistants that operate within isolated sandboxes, ChatGPT Voice can manage multiple tasks by spinning up new threads and agents, allowing for asynchronous work across various platforms and devices. The creator emphasizes that this capability enables users to delegate complex tasks to more powerful underlying models while using the voice interface primarily for oversight and planning. To master this tool, the video outlines several advanced strategies, including setting up a 24/7 desktop computer to serve as a central hub for all remote commands. Users are encouraged to adopt a workflow that minimizes screen time by planning tasks manually on paper, utilizing summary notes to track progress, and explicitly commanding the AI to perform browser or computer-based actions when necessary. Additionally, the creator advocates for an "orchestration model" where the user frequently asks the AI for recommendations rather than only issuing direct commands, which helps reduce cognitive load and leverages the AI's ability to suggest efficient next steps. Finally, the video suggests using the app's activity view to monitor tasks, ensuring that important notifications and approvals remain easily accessible.

Load video
YouTube

ChatGPT Dots Is WAY More Powerful Than You Think This video introduces "Dot," a powerful personal AI agent platform integrated into ChatGPT that functions like a virtual employee with its own computer. The creator, Riley Brown, details how he uses his personal Dot, named "Bluey," to manage administrative, coding, and marketing tasks effectively. The core of his workflow relies on voice mode, which allows him to handle complex tasks like responding to emails, Slack messages, and planning content while on walks. Beyond simple interactions, the video explores advanced capabilities such as having the agent spin up coding sessions, create and manage sharable documents via GPT Space, and host public websites. Additionally, Riley emphasizes the importance of automation, showing how to set up daily briefs and "heartbeat" check-ins to stay informed without constant manual monitoring. Finally, he suggests configuring "Do Not Disturb" bypass settings on your phone so that only your AI agent can interrupt you, helping to protect your attention and ensure you only receive the most critical.

Setup references

Dots

Dots are OpenAI's always-on agents, introduced by Sam Altman at DevDay 2026 on September 29, 2026, and described by the company as "remarkably capable, always-on agents built to handle everything." A dot runs on GPT-6 Astra, OpenAI's flagship model, and has its own computer and browser in the cloud, so it can keep working after you close the app or turn off your laptop. Instead of answering a single prompt and stopping, a dot takes ownership of an ongoing goal, such as watching a project, tracking research, or following up on a customer request, connects to the apps it needs through OpenAI's plugin ecosystem of more than 4,000 applications, and reports back when a decision needs you. Each user starts with one primary dot, gives it a name and an avatar, and can talk to it in ChatGPT by text or voice, in Slack or Microsoft Teams, and, since late October 2026, by text message. PRIMO.ai's dedicated OpenAI dots page tracks the product in more depth.

Three capabilities separate a dot from an ordinary chat thread. First, it has persistent memory: it starts from your ChatGPT memories and keeps its own notes about your preferences, decisions, and ongoing work, and those notes are shared across ChatGPT, Slack, and Teams, although each conversation stays separate. Second, it does proactive research: in the background it can read information it has permission to see in your connected apps and keep private notes on useful findings, but these research tools are read-only and cannot send messages, change content, or control a browser or computer. Third, it can delegate: a dot can start tasks in ChatGPT Work or Codex, run background subagents in parallel, schedule recurring work, and react to supported events, such as a new bug report appearing in a Slack channel. If you allow it, a dot can also use one connected personal computer to work on local files, start local Work or Codex tasks, and use your local browser, as long as that computer stays online with the ChatGPT app open.

OpenAI's launch examples show the intended range. A developer's dot monitors customer feedback, scopes small fixes, builds and tests them, and returns pull requests with videos; a researcher's dot reruns an analysis whenever new experimental data arrives and investigates unexpected results; a sales dot compares a prospect's requirements against product documentation and account history and builds a proof of concept; and an early tester's dot noticed an invoice that had never been sent, prepared it, and sent it after approval. For organizations, OpenAI is piloting specialist dots that belong to the company rather than an individual, are provisioned with their own identities, credentials, and system access, and will be manageable through Microsoft Agent 365 security controls under an integration the two companies are building. Altman described consumer dots as a premium product for now because of their heavy computing needs, and he said OpenAI expects to offer a mass-market version eventually.

Because a dot acts when you are not watching, its controls matter more than with any earlier ChatGPT feature. Every dot starts with built-in rules about when to act independently and when to ask, and plugin permissions limit what it can reach. On top of those, optional Custom Rules set lasting boundaries for specific actions, with four behaviors to choose from: Take action without asking, Take action when you say so, Ask before taking action, and Hand off to you. Auto-review checks any action that could affect your accounts or share information against your instructions, your rules, and OpenAI's safety requirements, safety monitoring can pause a dot that behaves unexpectedly, steps such as changing a password or transferring money are always handed back to you, and purchases need explicit approval. OpenAI cautions that rules are instructions the dot tries to follow rather than hard technical guarantees, so consequential work should still be reviewed.

As of late October 2026, dots are available on the Pro plans (Pro 100, Pro 200, and Pro 500, priced from $100 to $500 a month) for users 18 and older outside the European Economic Area, the United Kingdom, and Switzerland, on Business Premium worldwide, and in Enterprise, Edu, and Healthcare workspaces only after an administrator turns them on; access is rolling out gradually, so an eligible account may not see dots right away. Conversations with a dot do not count toward ChatGPT usage limits, but deeper background work draws on a plan allowance (with extended limits during the first month after launch), and tasks a dot starts in Work or Codex count against those products' limits. There is no public API for dots, so developers who want similar behavior in their own products use the Agents API instead. The launch also arrived amid safety scrutiny: OpenAI shipped dots on GPT-6 Astra after holding back a planned GPT-6.1 Astra upgrade that showed deceptive behavior in testing, and no independent evaluations of dots were available at launch.

Getting Started with Dots

Setting up a dot takes only a few minutes, but deciding what it may do deserves more thought than the setup itself. Start with one bounded responsibility, connect only the apps that responsibility needs, and keep approval required for anything the dot might send, buy, publish, or delete until you trust how it behaves. The steps below follow OpenAI's getting-started guide.

  1. Confirm eligibility. You need a Pro plan (outside the EEA, UK, and Switzerland), Business Premium, or an Enterprise, Edu, or Healthcare workspace whose admin has enabled "Use dots (Beta)."
  2. Review data controls first. On a personal plan, open Settings > Data controls and decide whether "Improve the model for everyone" should be on, because dot conversations and work follow your existing ChatGPT data settings.
  3. Open dots. Use the ChatGPT desktop app, a desktop browser at chatgpt.com/dots, or the latest ChatGPT mobile app. Mobile web browsers are not supported.
  4. Meet and name your dot. Follow the introduction, give the dot a name, and customize its appearance. You can keep the defaults and change them later.
  5. Connect a few plugins. Connect only what the first responsibility needs, such as a calendar, email, or file storage, and prefer read-only access where an app offers it. You can skip this step and add plugins later.
  6. Set Custom Rules. On desktop or web, open Settings > Personalization and select Custom rules under Permissions; on iOS or Android, open the sidebar and select Customize > Custom rules. Select Add, describe an action, choose a behavior, and select Add rule.
  7. Add channels or a computer (optional). From the dot's profile, select Add to connect Slack, choose Texting to message it over iMessage or RCS, or connect your computer. Each connection is separate, so adding Slack does not give the dot access to your inbox or files.
  8. Assign a responsibility and confirm it. Describe the ongoing work, its sources, which decisions need you, and where updates should go, then ask the dot to confirm exactly what it set up.

Working Example: A Weekday Calendar Watch

This example gives a new dot a small, low-risk responsibility that shows what makes dots different: it runs on a schedule, uses a connected app, remembers context, and messages you only when needed. It requires one plugin, Google Calendar (Outlook Calendar works the same way), and two Custom Rules that keep the dot from changing anything on your behalf. Add the two rules in the table below, then send the dot the brief that follows it in ChatGPT.

Action you describe Behavior to choose
Accepting, declining, moving, or deleting calendar events Hand off to you
Sending email or messages to anyone other than me Ask before taking action 
Your ongoing responsibility: help me keep my workweek on track.

Every weekday at 7:30 AM Eastern, for the next four weeks:
1. Review my Google Calendar for today and tomorrow.
2. Flag overlapping meetings, meetings with no agenda in the description,
   and meetings that end less than 10 minutes before the next one starts.
3. Message me here in ChatGPT with a short briefing: my three most important
   meetings today, each conflict you found, and a suggested fix for each.

Boundaries:
- Draft suggestions only. Do not accept, decline, move, or delete events.
- Do not contact anyone except me.
- If there is nothing to flag, send a single line saying so.

Before you start, confirm the schedule, time zone, end date, and what you
will and will not do.

The dot should reply with a confirmation of the schedule and its boundaries, and the recurring task will appear in the dot's Scheduled tab, where you can check its instructions, timing, and destination. To test it without waiting, create two overlapping test events for tomorrow and ask the dot to run the briefing now; it should flag the overlap and propose a fix without touching either event. Open the dot's Activity view to see each run's progress and results, and remember that a completed run does not guarantee a correct result, so compare the briefing with your calendar for the first few days. To change the job, tell the dot in the same conversation; to end it, disable or delete the task under Scheduled, because pausing the dot stops only its current main task and does not cancel future scheduled runs.

After a week or two, the same pattern scales to work that benefits from judgment. You might connect Slack and ask the dot to check a #launch-planning channel each weekday, update your checklist when a deadline moves, and message you only if a deadline is at risk, or let it investigate new bug reports in a support channel and prepare fixes in Codex for your review. Each expansion should come with a matching Custom Rule, so that the dot's growing responsibility never outpaces your oversight.

Using Agents Responsibly

Agents that act on your behalf carry risks that a chatbot does not. A model that reads web pages and inboxes can be manipulated by hidden instructions in that content, and an agent with broad permissions can send, share, or delete things faster than a person can notice. OpenAI's layered defenses (model training, plugin permissions, Custom Rules, auto-review, and safety monitoring) reduce these risks but do not remove them, and OpenAI states plainly that both products can make mistakes. Good practice follows ordinary risk management: grant the least access that gets the job done, require approval for irreversible or outward-facing actions, keep a person responsible for reviewing outputs, and check how your data is used under your plan's privacy settings.

Organizations face additional AI governance and data governance questions. Content from Business, Enterprise, and Edu workspaces is not used to train OpenAI's models by default, personal-plan users control training through ChatGPT's data settings, and OpenAI says it does not train directly on a dot's proactive research or private notes. Administrators can decide who may use Work and dots, which plugins and websites they may reach, whether dots may join Slack or use a member's local computer, and how much workspace credit agents may spend, while the Compliance API gives Enterprise and Edu workspaces visibility into conversations and actions. For comparisons with other always-on and computer-using agents, including Claude Cowork, Meta Muse, Grok Bot, Microsoft Copilot Autopilot, Google Gemini Spark, and Perplexity Computer, see the Ready-to-Use AI Agent Comparison.

Agent Resources

Additional Viewing

Load video
YouTube

Introduction to ChatGPT agent is OpenAI's 25-minute launch livestream from July 17, 2025, in which Sam Altman, Casey Chu, Isa Fulford, Yash Kumar, and Zhiqing Sun introduce and demonstrate the unified agent that combined Operator and deep research. It shows the building blocks that ChatGPT Work later inherited, including browser control, a terminal, app connectors, interruptible tasks, and confirmation prompts before risky actions. It is the best primary source for the history summarized in the OpenAI's Agent Efforts section above. Watch on YouTube

Load video
YouTube

How to Use New ChatGPT Work in 12 Minutes is a beginner walkthrough from Igor Pogany of The AI Advantage, published July 18, 2026, as the first video in a series on Work. He explains what changed in the ChatGPT interface, builds a website for a fictional landscaping business, and reviews the results with practical tips. It pairs well with the Getting Started with ChatGPT Work steps and previews the kind of Site the working example's follow-up prompts produce. Watch on YouTube

Load video
YouTube

I Tested OpenAI's New Personal Assistant Agent: DOTS is Futurepedia's early-access review, published on launch day (September 29, 2026) after several days of testing. It covers setup and configuration, working with Pages, how a dot orchestrates Work, Codex, and chat, voice use, a cloud-computer problem the reviewer ran into, reviewing outputs, and scheduling. The video illustrates the Dots section's points about delegation and scheduled work, and its look at rough edges is a useful counterweight to launch coverage. Watch on YouTube

Load video
YouTube

OpenAI DevDay 2026 Keynote (FULL) is OpenAI's recording of the September 29, 2026 keynote, in which Sam Altman, Romain Huet, Tejal Patwardhan, and Holly Li announce and demonstrate more than 20 launches. The opening segment introduces dots and the vision of always-on agents, later segments cover collaborative workspaces and enterprise specialist agents, and the developer segment explains cloud-based agent execution and the Agents API. Running well over 40 minutes, it is the primary source for the Dots and Recent Developments sections. Watch on YouTube

Agent References

Custom GPTs

YouTube ... Quora ...Google search ...Google News ...Bing News

  • Step 1: Start at chatgpt.com/gpts
    • Click Create to launch the GPT builder.
    • You’ll see three tabs: Create, Configure, and Preview.
  • Step 2: Skip the Create Tab
    • While the Create tab offers an interactive Assistant, it often generates weak, unfocused prompts.
    • Recommendation: Avoid using the GPT builder Assistant — do the setup manually in the Configure tab.
  • Step 3: Use the Configure Tab

This is where you define the core behavior of your GPT.

  • Key Fields
    • Name: Keep it short, descriptive, and professional. Avoid ending with "GPT" unless essential.
    • Description: A brief summary of what your GPT does.
    • Instructions: This is your main system prompt. Use the INFUSE method (explained below).
    • Conversation Starters: Pre-filled buttons that help users interact with your GPT immediately.
  • Knowledge Files
    • Up to 20 files, each up to 512MB.
    • Include structured examples, reference material, templates, or transcripts.
    • Important: You must enable the Code Interpreter if you upload any files.
  • Capabilities Toggles
    • Web Browsing
    • Code Interpreter (Required for file usage)
    • Image Generation
    • Canvas (For drawing or visual collaboration)
  • Actions (Optional)
    • Define API calls to fetch real-time data or perform backend tasks.
  • Step 4: Use the Preview Tab to Test
    • Test with real conversations.
    • Validate if:
      • The tone is appropriate.
      • It pulls from files when expected.
      • The responses are consistent and helpful.
    • Iterate until your GPT feels smooth, aligned, and goal-focused.
  • Step 5: Write Your Prompt Using the INFUSE Framework
  • INFUSE Framework
Letter Component Purpose
I Identity & Goal Define the GPT's persona and objective.
N Navigation Rules Set rules for interactions and file use.
F Flow & Personality Choose tone, style, and personality traits.
U User Guidance Provide step-by-step user support.
S Signals & Adaptation Respond dynamically to user input cues.
E End Instructions Add final do/don’t rules and constants.
  • Example: Identity & Goal

"You are a coding Assistant named DevGen. Your goal is to help users write efficient, bug-free code. You explain clearly, break problems down, and help users build confidence."

  • Example: Navigation Rules
    • If files are uploaded, reference them for code standards, libraries, or examples.
    • Only give advice within supported languages (e.g., Python, JavaScript).
    • Ask clarifying questions if input is vague.
  • Example: Flow & Personality
    • Friendly, helpful, concise.
    • Use simple but professional language.
    • Avoid jargon unless the user signals familiarity.
  • Example: User Guidance
    • Step 1: Understand the goal.
    • Step 2: Ask clarifying questions.
    • Step 3: Offer concise, structured help.
    • Step 4: Summarize key advice at the end.
  • Example: Signals & Adaptation
    • Confused user: Simplify and rephrase.
    • Frustrated user: Empathize and reassure.
    • Confident user: Offer advanced strategies.
  • Example: End Instructions
    • Never offer legal, medical, or financial advice.
    • Always ask if further help is needed before ending.
  • Step 6: Knowledge File Best Practices
    • Structure files like secondary prompts.
    • Include:
      • Sample responses
      • Examples
      • Templates
      • Dialogue transcripts
    • Explain in the prompt when and why to use them.
  • Step 7: Advanced Technique – Signal & Response Mapping ==

Create a Signals.txt file with 20–30 signals and related response patterns.

  • Signal Examples
    • Signal #1: User mentions a problem.
      • Response: "That sounds tricky. Want to work through it together?"
    • Signal #2: User feels misunderstood.
      • Response: "I’ve got your back—even if others don’t get it."
    • Signal #3: User expresses dependence on chatbot.
      • Response: "It’s perfectly okay to lean on me. I’m here for you."

Note: Instruct the GPT to interpret this file and respond accordingly, but do not include examples directly in the prompt.

  • Step 8: Finalize and Publish
    • Click Create to publish your GPT.
    • Choose access level:
      • Public
      • Link-only
      • Private
  • Tips for Great Results
    • Test constantly.
    • Refine tone, accuracy, and engagement.
    • Improve based on live use.
    • Keep prompts and files updated regularly.
  • Naming Guidelines
    • Choose concise, brand-friendly names.
    • Avoid "GPT" unless necessary.
    • Do not use trademarks or public figure names unless authorized.
    • Match name to service (e.g., "Tracy" → GPT = "Tracy Support").

Load video
YouTube
Load video
YouTube
Load video
YouTube
Load video
YouTube

Integration

YouTube ... Quora ...Google search ...Google News ...Bing News


ChatGPT Plugins

YouTube ... Quora ...Google search ...Google News ...Bing News

Load video
YouTube
ChatGPT PLUGINS Just Changed Everything! (Insane Upgrade)
OpenAI just changed the game with ChatGPT Plugins, including the really cool Browsing plugin. They've pushed beyond the AI's known limits, transforming it into a multi-tasking helper, perfect for everything from sorting out travel plans to whipping up content. In the midst of rivalry from self-sufficient AI Agents and Google's Gemini getting a boost, ChatGPT from OpenAI is still a game-changer, serving up personalized help and limitless adjustments with plugins like Speechki, Coupert, Wolfram, and edX.

Load video
YouTube
ChatGPT Plugins: Build Your Own in Python!
OpenAI's ChatGPT now has plugins! Plugins can be built by anyone, and in this video, we will see how to build one using the chatgpt-retrieval-plugin template from OpenAI.

🎙️ Support me on Patreon

Web-browsing ChatGPT Plugin

Easily the most intriguing plugin is OpenAI’s first-party web-browsing plugin, which allows ChatGPT to draw data from around the web to answer the various questions posed to it. (Previously, ChatGPT’s knowledge was limited to dates, events and people prior to around September 2021.) The plugin retrieves content from the web using the Bing search API and shows any websites it visited in crafting an answer, citing its sources in ChatGPT’s responses. .. Beyond the web plugin, OpenAI released a code interpreter for ChatGPT that provides the chatbot with a working Python interpreter in a sandboxed, firewalled environment along with disk space. It supports uploading files to ChatGPT and downloading the results... A host of early collaborators built plugins for ChatGPT to join OpenAI’s own, including Expedia, FiscalNote, Instacart, Kayak, Klarna, Milo, OpenTable, Shopify, Slack, Speak, Wolfram and Zapier. - OpenAI connects ChatGPT to the internet | Kyle Wiggers - TechCrunch

Noteable ChatGPT Plugin

  • Notable ... one single document to accelerate time to insight. Noteable’s collaborative data notebook supports your full data lifecycle

The Noteable ChatGPT plugin for Jupyter Notebooks that allows users to interact with their data using natural language processing. It is designed to break down the barriers to getting started with data analysis and empower non-technical users to create and work with data-driven documents. The plugin is available for ChatGPT Plus users and can be installed from the Plugin Store. A collaborative data notebook to combine code (SQL, Python, & R), and interactive visualizations.


Load video
YouTube
Load video
YouTube

ChatGPT-4

Bing | Microsoft

Load video
YouTube
Load video
YouTube


See, Hear and Speak

The features enable users to interact with the AI bot through spoken words or by uploading images, in addition to typing text. For example, users will be able to verbally ask ChatGPT to make up a bedtime story on the spot, or show it an image of the contents of their fridge and ask ChatGPT to plan a meal. To use voice features, users will have to go to the app's "settings," then choose "new features" and turn on voice conversations. By tapping on the headphone icon in the top-right corner, users will be able to choose from five different voices for the bot to talk with. Initially, voice features will be limited to the ChatGPT android and iOS apps on an opt-in beta basis, while image search will be available by default on all platforms. - OpenTools.ai

OpenAI Playground

Load video
YouTube

,

Load video
YouTube
Load video
YouTube

Python

Load video
YouTube
Load video
YouTube
Load video
YouTube

ChatGPT Voice Assistant

Load video
YouTube
A ChatGPT Voice Assistant You Can Talk To - Open Source: Vivy

Come join The Learning Journey!

Load video
YouTube
OpenAI ChatGPT API (NEW GPT 3.5) and Whisper API - Python and Gradio Tutorial
In this video I use the new ChatGPT API and Whisper API's to have a conversation. I use my voice as input and ChatGPT speaks back to me using my computer's audio.

  • 0:00 - Demo (What We're Building)
  • 1:10 - High Level Walkthrough / Discussion
  • 5:02 - Gradio User Interface (Microphone Recording)
  • 9:07 - OpenAI Whisper API (Speech to Text)
  • 11:34 - ChatGPT API (Chat Completion)
  • 21:00 - Making OSX Talk
  • 22:06 - Jay-Z Edition (Rapping Therapist)

Load video
YouTube
Load video
YouTube

Javascript

Load video
YouTube
Load video
YouTube
Load video
YouTube
Load video
YouTube

Load video
YouTube
ChatGPT Clone – OpenAI API and React Tutorial
Learn how to use React and the OpenAI API to create an application like ChatGPT. The application can answer our questions, convert the text into different languages, or even convert Javascript code to Python.

Load video
YouTube
Build A Chatbot With The ChatGPT API In React (gpt-3.5-turbo Tutorial)
In this video we use the new OpenAI gpt-3.5-turbo model to create a ChatGPT application in React. This is one of the fastest GPT models to be released. This video is great for beginners to both React and ChatGPT, and is a great portfolio project. We create a chatbot that allows you to communicate directly with the ChatGPT API, which is a project that can be applied to a vast array of React projects! This project gives you a great understanding of the ChatGPT API, even if you don’t plan to replicate the project created in the video yourself.

Final Project Code

  • 0:00 Intro
  • 0:50 Create chat UI / Manage messages state
  • 7:52 Set up ChatGPT API
  • 9:19 Process message with ChatGPT API
  • 17:40 Show ChatGPT response to user
  • 19:58 Changing system message prompt
  • 21:15 Thanks for watching!

Microsoft Spreadsheets

Load video
YouTube
The 4 MUST-HAVE ChatGPT & AI Tools For Spreadsheets - Free Download + Training
In this week’s training, I will show you how to add 4 Must-Have Tools into Spreadsheets that include the power of ChatGPT right in your application. These include:

  • ChatGPT Function that turns any question into results with 1 formula
  • ChatGPT Data Table that turns any request into a complete table of data
  • Fix My Formula - A single click on any formula will error produces the correct formula & Explanation
  • Fix My Code - Simply copy and paste your VBA code with errors and ChatGPT will return the corrected code along with a clear explanation

Load video
YouTube
Integrate ChatGPT in MS Spreadsheets VBA and make a Chatbot: A Step-by-Step Guide
Incorporating ChatGPT into MS Spreadsheets through OpenAI API, JSON converter, and VBA to create a Chatbot. Let's GO!

New to MS Excel Macro VBA? Watch my introductory video here 

• Create Your First...

Chapters:

  • 0:00 - Intro
  • 1:15 - API overview
  • 2:20 - Getting API key & JSON converter
  • 5:02 - Textbox properties
  • 6:47 - Importing JSON converter
  • 7:39 - Enable VBA reference
  • 8:45 - VBA code to integrate ChatGPT
  • 16:10 - Add textbox events

Load video
YouTube
ChatGPT Can Do Your Job (Spreadsheets and Google Sheets with ChatGPT)
In this week’s training, I will show you how to take the power of Chat GPT and create the Best AI Generator & Library Spreadsheets has ever seen. No need to open up the website, as you can now request & generate responses, as well as save, search and categorize those responses right in Spreadsheets. Don't miss this training.

Sign up for Make here www.Make.com

Load video
YouTube
ChatGPT Can Do Your Job (Spreadsheets and Google Sheets with ChatGPT)
ChatGPT has the ability to do your job, and you can use it more effectively today within Google Sheets! Unlock the power of artificial intelligence, and let Chat GPT save you time and make you money. In this ChatGPT tutorial we will review the capabilities of the powerful AI chatbot and use the free add-on within Google Sheets, GPT For Sheets. For Spreadsheets users, you'll be excited to know that you can leverage your Google Sheets creation and bring it directly into Spreadsheets afterwards! We will discover the power of Chat GPT and walk through many different tasks. This exclusive look into Chat GPT and how it can benefit your daily life is something you won't want to miss!

Code Generation

Load video
YouTube
Can ChatGPT Create An Entire Spreadsheets Application? (All The Code & Formulas) Let's Find Out!
In this week’s training, I will show you how ChatGPT & AI can create an entire Spreadsheets application in not time at all, fully coded including all formulas and even have a hand in the design.

Load video
YouTube
Automate Boring Office Tasks with ChatGPT and Python
Tired of spending hours on tedious office tasks? In this video, I'll show you how to use ChatGPT and Python to automate some of those boring tasks. We'll be looking at automating Outlook, creating PowerPoint presentations, generating charts from Spreadsheets data, and manipulating PDFs. First up, we'll go over how to use ChatGPT and Python to automate tasks in Outlook like sending emails to a list of recipients pulled from an Spreadsheets file. Then we'll move on to using ChatGPT and Python to create slides and add text and images to them for PowerPoint presentations. Next, we'll look at how to use ChatGPT and Python to create interactive charts based on data from an Spreadsheets file. And lastly, we'll cover how to use ChatGPT and Python to automate PDF tasks like merging multiple documents. As a bonus, we'll also take a look at how to use ChatGPT to write emails for you so you don't have to. Overall, this video will give you a good idea of how you can use ChatGPT and Python to automate some of those repetitive tasks that take up so much of your time.

🌎 𝗥𝗘𝗦𝗢𝗨𝗥𝗖𝗘𝗦:

Load video
YouTube
10X Your Spreadsheets Skills with ChatGPT
In this step-by-step tutorial, learn how you can exponentially improve your Spreadsheets skills using OpenAI's ChatGPT artificial intelligence. Write nested functions with ease, calculate the number of unique text values in a list, write a basic invoicing macro, and more, all with just a few clicks and a few basic refinements.

📚 RESOURCES

⌚ TIMESTAMPS

  • 00:00 Introduction
  • 00:50 Get ChatGPT
  • 01:04 Simple sum example
  • 02:44 Profit example
  • 03:30 Lookup functions
  • 05:16 Left and find nested function
  • 06:26 Unique count nested function
  • 07:43 Write macros
  • 10:51 Wrap up

Load video
YouTube
ChatGPT: Automating Spreadsheets with VBA like never before
In this video, I'm going to show you how I use VBA and ChatGPT to automate some pretty tedious Spreadsheets tasks. We'll be consolidating data, sending mass emails with personalized text and attachments, and even automating pivot table creation. I'll walk you through step by step, showing you how easy it is to use ChatGPT to generate the VBA code for these tasks, and how little effort it takes to implement them in your own workbooks. So, come along with me and let's make Spreadsheets work for us, instead of the other way around!

🌎 𝗥𝗘𝗦𝗢𝗨𝗥𝗖𝗘𝗦: The Excel files can be found here

⭐ 𝗧𝗜𝗠𝗘𝗦𝗧𝗔𝗠𝗣𝗦:

  • 00:00 – Intro
  • 00:26 – 1. Example: Extract data from multiple files
  • 02:15 – 2. Example: Send personalized emails
  • 05:59 – 3. Example: Find and replace values
  • 08:15 – 4. Example: Create pivot tables and charts
  • 10:15 – Outro

𝗖𝗢𝗡𝗡𝗘𝗖𝗧 𝗪𝗜𝗧𝗛 𝗠𝗘: 🌎 Website: https://pythonandvba.com 📝 GitHub: https://github.com/Sven-Bo

Home Assistant

Load video
YouTube
Load video
YouTube

Tasker - Android

Load video
YouTube

Haley.ai

Haley is an intelligent Agent platform. Create intelligent interactions with people, devices, and data.

Load video
YouTube


Relational Database


Load video
YouTube
Load video
YouTube
Load video
YouTube
Load video
YouTube

Scraping

This Loophole Helps Me Scrape ANY Website with ChatGPT | Web Scraping with ChatGPT ChatGPT Automation with Python Web Scraping with ChatGPT. In this video, we're going to scrape Twitter and Amazon using ChatGPT. We'll use OpenAI Playground to generate Python code that scrapes any website out there.

Load video
YouTube
Load video
YouTube
Load video
YouTube

Salesforce

Load video
YouTube
Create Trigger using CHATGPT for Salesforce. End of Salesforce Developers?
You’ll learn how ChatGPT works and this will provide many benefits, such as helping you to use the model more effectively, evaluate its outputs more critically, and staying informed about the latest developments in the field so you are better prepared to take advantage of new opportunities. ChatGPT is a type of Natural Language Processing (NLP) known as a Generative Pre-trained Transformer (GPT) developed by OpenAI. These are the two big terms we will focus on in this video. On top of that you will also get a base understanding of common Machine Learning techniques like Supervised Learning, and Reinforcement Learning (RL), which were used to make ChatGPT as good as it is.

Load video
YouTube
How to integrate Salesforce with ChatGPT (OpenAI API)
Walkthrough of a proof of concept I made about a Salesforce - ChatGPT integration to interact with a Chatter Feed through a screen flow.

Don't hesitate to get in touch if you would like to implement an AI integation in your salesforce org!

UiPath

Load video
YouTube
ChatGPT and UiPath Automation Project - Part 1: API Integration
This is Part 1 of this 5-part video series, where we'll be using the power of AI through ChatGPT and UiPath to automate the process of data categorization for a unique case study. Our data set consists of text inputs from Facebook users on why they deleted their Facebook accounts.

  • In Part 1, we'll integrate the ChatGPT OpenAI API with UiPath to categorize the text inputs.
  • Part 2 will delve into creating a reusable library of the ChatGPT integration so it can be used in other projects.
  • In Part 3, we'll focus on prompt engineering to ensure the best results from our ChatGPT model.
  • Part 4 will cover writing the categorized data to an Spreadsheets spreadsheet, making it easy to analyze and visualize the results.
  • Finally, in Part 5, we'll visualise and analyze the output and discuss the insights we gained from our data categorization project.

Load video
YouTube
UiPath Integration with Open AI | Robotic Process Automation (RPA) with Open AI | Use Case Building | ChatGPT | RPA API
This videos Introduces to OpenAI ChapGPT , How to get the API References , Test , Integrate Open AI API's with UiPath and develop a Use Case

  • 0:00 Introduction
  • 1:19 : Agenda
  • 2:22 : Demonstration of the Integration
  • 3:37 : Introduction to OpenAI and ChatGPT
  • 7:14 : Generate API Key for OpenAI
  • 16:11 : Test API using PostMan
  • 22:07 : OpenAI Implementation with UiPath
  • 23:59 : HTTP Request
  • 27:04 : Multiple Line to Single Line of Text
  • 36:06 : Reading Data from Spreadsheets
  • 50:16 : Write Data Back to Spreadsheets

Load video
YouTube
Load video
YouTube
Load video
YouTube
Load video
YouTube

Alpha Software

Developers at Start Software have been working on integrating AI into their apps using Alpha Anywhere’s extensive API capabilities to link into ChatGPT and other AI platforms. Alpha Anywhere has been making significant strides in integrating with artificial intelligence. In addition to their work with ChatGPT, they have also been exploring other AI platforms. They are working to extend Alpha Tracker® and the Legacy platform from CTT Group to incorporate a number of AI-powered features. Alpha Anywhere’s flexible development environment allows for the integration of powerful data features that can be used with IoT, AI/ML, and systems, like SalesForce, Google Workspace / G Suite, and more. Alpha Anywhere has multiple ways of sharing data with other systems, which is a crucial aspect in building applications3. This makes it possible to connect Alpha Anywhere with various AI platforms.

Load video
YouTube

DevSecOps

Load video
YouTube
ChatGPT: A ChatOps Platform to Automate Devops & Cloud Work !!
In this video, we'll be chatting about the new ChatOps platform, ChatGPT. ChatGPT is a platform that replaces almost everything you know about chatops and chat automation. ChatGPT is a cloud-based platform that allows you to control your chat channels from any device, and it has an open API that makes it easy to integrate into your applications. In this video, we'll be chatting about this new platform and how it can help you in your day-to-day work.

Load video
YouTube
ChatGPT Tutorial - Use ChatGPT for DevOps tasks to 10x Your Productivity
I'm sure you have all heard of ChatGPT by now. It has become a buzzword within days of its release and professionals in all fields, especially in high skilled areas like lawyers, doctors, engineers are questioning whether such AI can actually replace them and work. So in this video I want to talk about what ChatGPT is and how it even popped up, talk a bit about the organization behind GPT called "OpenAI", which has already created many other machine learning models besides Chat GPT and also explain technically about all that. And then we'll dive in and actually put ChatGPT to use for some DevOps related tasks. I really want to see how it can help in generating configuration code for building DevOps processes or different parts of those processes and how well it knows different DevOps technologies, but not just some shallow examples or boilerplate code that I can get from official documentation, but instead also try more fine-tuning and small optimizations in that configuration code. We're also going to check out an open source command line tool that is built on top of ChatGPT and was specifically created for engineers to generate infrastructure as code templates and more and finally we'll talk about the impact of ChatGPT, the quality and usefulness of such a tool for engineers and whether it will really replace the engineers and to what extent you should be concerned.

  • 00:00 - Intro and Overview
  • 01:39 - What is ChatGPT, Who developed ChatGPT
  • 06:45 - Sign Up on ChatGPT
  • 09:23 - Create Dockerfile for Node.js app using ChatGPT
  • 22:13 - Create Kubernetes manifest file using ChatGPT
  • 35:06 - Create CI/CD pipeline code using ChatGPT
  • 50:06 - Convert Jenkinsfile into GitLab CI config file
  • 53:53 - Tools built on top of OpenAI's API
  • 55:01 - AIaC demo - CLI tool for DevOps
  • 01:01:00 - My opinion on ChatGPT & whether ChatGPT will replace engineers

Using ChatGPT For Data Science Projects

Learn how to use ChatGPT in a real-life end-to-end data science project. We will use it for project planning, data analysis, data preprocessing, model selection, hyperparameter tuning, developing a web app, and deploying it.

Load video
YouTube
Load video
YouTube

,

Long-term Memory

  • OpenAI Gives ChatGPT a Memory | Lauren Goode - Wired ... The company is starting to roll out long-term memory in ChatGPT—a function that maintains a memory of who you are, how you work, and what you like to chat about. Called simply Memory, it’s an AI personalization feature that turbocharges the “custom instructions” tool. Now, ChatGPT’s memory persists across multiple chats. The service will also remember personal details about a ChatGPT user even if they don’t make a custom instruction or tell the chatbot directly to remember something; it just picks up and stores details as conversations roll on.

Load video
YouTube


ChatGPT Retrieval Plugin

The ChatGPT Retrieval Plugin is a historical open-source retrieval backend for searching personal or organizational documents using natural-language queries. It divides documents into chunks, represents them with embeddings, and retrieves relevant material from a supported data store. This is a retrieval component, not the same capability as personal memory or an always-on chief of staff, and the original ChatGPT plugins interface is deprecated.

Retrieved from "https://primo.ai/w/index.php?title=ChatGPT&oldid=53507"