Couldn’t I always ask LSBot to review an assignment with me?

Yes. You can type something like, “Review this assignment with me,” or “Quiz me on the concepts covered here.” LSBot handles those requests well, but it typically asks you to steer the conversation.

With the new End-of-Assignment Review, LSBot takes the lead. You respond to a curated sequence of questions based on the concepts covered in that assignment.

When you start a review, LSBot generates one question at a time. Question types may include:

• Conceptual Questions
  “Can you explain the difference between a shallow and deep copy, along with an example?”
• Reading Code
  “Explain what happens when we run this code. What is the output? What concept is being demonstrated?”
• Targeted Quiz
  “Here are 10 values. Identify each as truthy or falsy and briefly explain your reasoning.”
• Debugging
  “Here’s a code snippet that isn’t working as expected. Identify and fix the error. What misconception likely led to this issue?”
• Coding Tasks
  “Write a function that takes two arguments and…”

For this feature, LSBot is designed to maintain forward momentum. It avoids getting sidetracked or offering excessive critique. You’re always free to provide additional direction, skip questions, or end the review at any time.

Where to find End-of-Assignment Reviews

We’ve just begun rolling out this feature. At the moment, it’s available in 101 and 110 courses.

Look for assignments with a chat icon next to the name.

At the end of those assignments, you’ll see a box describing the End-of-Assignment Review and a button to start.

You can complete a review as many times as you like and answer as many or as few questions as you prefer.

We’d love your feedback. Please try it out and let us know whether you find it valuable and whether we should expand it across the curriculum.

End-of-Assignment Reviews with LSBot was originally published in Launch School on Medium, where people are continuing the conversation by highlighting and responding to this story.

A couple of weeks ago, we introduced LSBot-Guided Reviews to help students prepare for assessments at Launch School. After receiving positive feedback, we decided to expand this new feature to support more use cases and reach a wider audience. Now, we’re excited to share that LSBot-Guided Reviews are available for free on the Open Bookshelf!

LSBot-Guided Reviews are an interactive collection of prompts designed to leverage AI to help you review and retain content. This feature is free and available now without signing in. Use them to reinforce what you know or to uncover gaps that need more focused study.

The Goal

One common challenge students report is uncertainty around how to use LSBot effectively for studying. Many rely on it mainly to explain concepts, which can unintentionally lead to passive learning: consuming explanations without deeply engaging with the material.

LSBot-Guided Reviews are designed to change that. They model active study strategies by putting the cognitive work back on you. You’ll be prompted to:

• Explain concepts to LSBot and receive feedback on your explanations
• Generate code snippets to quiz yourself
• Compare and contrast related concepts to develop overarching mental models

..and generally improve your ability to use AI as a tool to evaluate your understanding, rather than replace it

As a reminder, all materials on the Open Bookshelf are free and publicly accessible, regardless of subscription status.

See it in action

Here’s a short preview showing how to access and use LSBot-Guided Reviews in the books:

Try it Now

LSBot-guided reviews are available in these books:

• Introduction to Programming with Python
• Introduction to Programming with Ruby
• Introduction to Programming with JavaScript

Since this is a new feature, we’re eager to hear your thoughts. What’s helpful? What’s confusing? How could we make it even better? Share your feedback with us at hello@launchschool.com. We’d love to hear from you!

LSBot-Guided Review for Books was originally published in Launch School on Medium, where people are continuing the conversation by highlighting and responding to this story.

Hi, everyone! Chris here.

I’ve been a bit late on posting 2024 salary data, mostly because I’ve already shared results from the 3 cohorts (see 2401, 2405, and 2408). But it’s overdue, and it’s nice to see the year’s data in one place. There have been a few updates since I last posted those cohort numbers. For example, all three internships converted, so it slightly improved our 6-month placement results, some folks from earlier cohorts have landed after 6 months, so I’ve included those in the salary averages, and finally, I get to share in-progress updates from 2025 cohorts.

Quick reminder that even though it’s “2024” salary data, a significant percentage of the job offers occurred in 2025. Here, 2024 signifies the start of each cohort and includes the 3 cohorts that began their Capstone participation sometime during 2024. For example, the fall 2024 cohort didn’t start their job hunt until Jan 2025. So even though it says “2024”, the placement numbers are still relevant now.

Note: We hosted a webinar after this data was published to share further insights and answer student questions. Once you’ve read this article, check it out here.

Year 2024 Totals

The “No Job Hunt” column includes participants who either didn’t finish Capstone or didn’t do a job hunt (either planned or unplanned). The “No Job (so far)” column includes participants who have paused their job hunt, are actively job hunting, or are not replying to us.

Placement Results

In the tables below, the “enrollees” column signifies the denominator. For example, 2401’s 6-month place of 69% for “enrollees” was calculated from 22/32. The “remove No Job Hunt” column shows the percent if we only include those who job hunted in the denominator.

Accepted within 6 months:

All placements, including those who accepted after 6 months:

US Salary Data

Out of the 56 job offers, 44 were in the United States. The numbers are in USD and represent base salary only.

Salary results remain strong despite the tighter market, which reflects continued demand for high quality job candidates. If you want to see individual cohort salary data, see 2401, 2405, and 2408 Reddit posts. If you want to compare with previous Capstone salaries, see launchschool.com/salaries.

Remote Jobs

We only track remote percentages for US placements.

No major difference in remote salaries due to the small number of SFBA/NYC salaries during 2024.

Despite the minor salary difference, there’s a palpable downward trend in remote job salaries. Compare the below with year 2022 and 2023 remote percentages of 78% and 73%, respectively.

What I’m Seeing on the Market

I wrote some notes 4 months ago when I posted cohort 2408’s numbers. I still agree with them, so I’ll copy/paste them here. After that, I’ve added some new notes for this post.

Copied from 4 months ago:

• No surprise, placement rates are getting worse and it’s taking longer to land jobs. Pleasantly surprised salary numbers are still solid.
• Folks are now spending more time in OSI/internships, which means a 6-month job hunt can take 8+ months.
• We’re now advising Canadian Capstone participants to aim for US-based startups; there are far fewer employers hiring at the entry/intermediate level in Canada.
• Job hunting is a skill onto itself. There is some correlation between technical prowess and how quickly someone lands, but there are too many technically capable people who are not skilled at job hunting. Ironically, these are exactly the type of people employers should be going after, if only they can find them.
• Depressed market causes two reactions for job hunters: a) aggression to meet the challenge or b) avoidance. It’s my observation that thoughtful, introverted people tend to choose the latter. The exact skills that allows one to be good at technical work ends up hurting in this type of market.

Additional notes:

• The biggest headwind Capstone candidates face right now is tech’s Return To Office (RTO), which is making remote job hunting very difficult. Most Capstone participants are still looking for remote work and that’s been the biggest issue affecting placement.
• We first introduced AI Engineering topics to cohort 2501, meaning the 2024 cohorts didn’t get this information. We ran a course backfilling AI Engineering curriculum for 2024 job seekers, which has been impactful. Several people landed opportunities specifically due to this addition.
• Capstone graduates are able to get interviews, but they’re more difficult to close. Interviews are more demanding, even if the exercises and take-homes aren’t more difficult, the judging is harsher.
• Using AI-assisted coding tools is becoming the standard, particularly among startups. Interviews are now assessing your fluency and preferences with these tools. Some take-homes now require you to submit the entire chatlog conversation used generate the solution.
• Technical skills are a starting point now. Companies are looking for high agency, leadership ability, superior communication, and in the case of some SF startups, 60+ hours per week in office.
• All that being said, our salaries still remain strong. In fact, some of our highest salaries of all time have come just in the past few months, including our very first base salary over $200k. Companies can’t find enough good people, but the bar for “good” has increased.
• Lower salary (<$70k) roles are nearly non-existent. The entire lower tier is now gone.

Future Looking

• SFBA and NYC will dominate job opportunities in 2026. For example, our two 2025 cohorts, 2501 and 2505, are both on the market. So far, 2501’s average salary is $94,846 while 2505’s average is $140,375. The difference: a whopping 62% of 2505’s accepted offers were in SFBA or NYC, while it was only 10% for 2501.

• RTO will continue through 2026. 2501 is currently at 17% remote placement while 2505 is at 20%. Here’s a graph of percent remote placement for previous years and cohorts.

• AI-assisted coding tools will be mandatory going forward. Mastering their fluency and developing expertise around this ecosystem will be critical. In particular, the use of agents and autonomous workflows will be important. Communicating precisely, performing code reviews, experimenting with architectures, managing junior devs, etc., were all tasks previously in the domain of team leads and more senior developers. But these skills are in high demand right now because they are required to work with AI-assisted coding tools. (If you’re in Core reading this, our advice is to learn the fundamentals, e.g., Core, without using these tools. You can more easily join the party once you finish Core)

Conclusion

• The Capstone formula still works, though it’s gotten harder. Our most optimistic placement metric sits at 82%, and that number will climb as more students land jobs. Still, that means roughly a fifth of people haven’t landed yet. Whether that’s good or bad in this market is tough to say. No one shares their placement data, not other bootcamps, not universities, not anyone. It’s definitely worse than our historical Capstone numbers, but I still believe there’s something special about the Core + Capstone approach. And from what I can tell, our results appear to be industry leading for 2024.
• The time of the career change tourist is over. Software engineering is no longer the obvious fallback for people who don’t know what to do or who just want an easy paycheck. The landscape is more competitive now. But for those who genuinely love it, there’s never been a better time to enter the field. New workflows and abstractions are being created in real time. There are few experts and no tenured elders telling you to “pay your dues”. This kind of chaotic disruption has always been tech’s real appeal. Those high software engineering salaries are a byproduct of that innovative disruption. Come for the salaries, stay for the world-changing opportunities. Just make sure you learn the fundamentals first.

Want to dive deeper? Watch the webinar where we discussed this data in detail and answered student questions:

Happy New Year!

2024 Capstone Salary Data was originally published in Launch School on Medium, where people are continuing the conversation by highlighting and responding to this story.

We’re excited to share several updates to LSBot and Launch School!

• New exercise type: Code Review Exercises — practice reviewing code you didn’t write, a crucial but under-practiced skill.
• Conversation threads — manage multiple threads to stay organized and clear stale chat history.
• Improved LSBot hints — more supportive, flexible, and aligned with your learning needs.

Code Review Exercises

You’ll now see some exercise groups labeled Code Reviews. What does that mean? Not that you get code reviews — LSBot can already provide reviews for all of your exercise solutions. Instead, these are exercises where you’re given both an exercise prompt and a solution written by someone else. Your task: review the solution for correctness, clarity, and best practices.

Why now?

Reading and understanding code is a huge part of software engineering. Overwhelmingly, developers spend more time reading code than writing it. If it’s so important, why did we wait until now to add these types of exercises? The challenge with code review exercises is that there’s no definitive solution. When we ask you to write a function that completes some task, we can give you test cases so that you know when you’ve solved the problem. That’s not so simple with a code review. Thanks to LSBot, we can now provide feedback on student-written code reviews.

While LSBot has unlocked this capability, there’s another aspect that makes these types of exercises even more critical now — AI-assisted development. Increasingly, developers read and evaluate code generated by AI models to decide whether or not to use it. Even more than before, the importance of parsing code is a fundamental skill.

Looking forward

For now, you’ll find code review exercises for PY101, JS101, and RB101. We’ll expand across the curriculum based on your feedback. This way, you can point out the N+1 query in LS180, or suggest using event delegation in JS230. Even if you aren’t in a 101 course, check out the exercises and let us know what you think!

Conversation Threads

In all LSBot Conversations, you can now manage multiple conversation threads. There are several use cases for this feature. You can use new threads as a “reset” feature to clear old messages, especially if LSBot made an error. You can also use threads to organize your conversations. If you’ve given LSBot specific instructions for quizzing you or generating exercises, you might want to use a different thread to search the curriculum or ask a conceptual question so that you keep those instructions focused. Read more about this feature and see a demo in the LSBot User Guide.

Improved LSBot Hints

Over the past few months, we have noticed that LSBot can sometimes be overly strict and not always respond in the way students need. Our goal is for LSBot to support your learning process, not limit it. So we’ve refined how hints work to make them more supportive and flexible.

If you tried this feature before and weren’t impressed, we encourage you to give it another shot. The interface remains the same, but under the hood, we’ve made significant improvements based on user feedback. In fact, the same goes for any LSBot feature — if it didn’t feel great the first time, it may be worth trying again. We’ve made updates across the board in book, assignment, and exercise features.

We’re happy to continue building out LSBot to be an always-on study aid for students. If you have any feedback, especially about our new features, let us know!

Code Review Exercises and other LSBot Updates was originally published in Launch School on Medium, where people are continuing the conversation by highlighting and responding to this story.

We’re excited to share a few new LSBot updates:

• LSBot Questions — now present in all Open Book Shelf books
• LSBot User Guide — upgraded into a book for easier navigation and reference
• Deep Dive Mode — a brand new option in LSBot Conversations

LSBot Questions

You’ve probably noticed LSBot Question Boxes throughout the curriculum. Now, they’re also available in every book on the Open Book Shelf.

LSBot Questions have been a huge hit with students — whether used for reflection on a first pass through the material or for quick self-assessment before an interview. We’re excited to make them publicly available so that non–Launch School students can try out this feature as well.

LSBot on the Open Book Shelf

The LSBot User Guide has leveled up — from a single document into a full book on the Open Book Shelf. That means easier navigation, smoother onboarding for new students, and a reliable place to check in whenever you’re curious about a feature. Read it here.

A bonus: because LSBot can now access this book, you can ask LSBot about LSBot. It can point you directly to the right chapter — for example, how to request a code review or how LSBot-graded quizzes work.

Deep Dive Mode

LSBot Conversations in the Launch School app now feature a new option: Deep Dive Mode. In short, this gives you more control over the tradeoffs that shape LSBot’s behavior.

Until now, LSBot conversations have emphasized speed and brevity: quick answers to keep you moving. With Deep Dive Mode, you can choose depth instead — longer, more detailed responses that draw on the full breadth of Launch School resources.

Deep Dive vs. Fast Mode: A Quick Comparison

Note: Before this update, LSBot conversations were in what we now refer to as Fast Mode. It just didn’t have a name yet because it was the only option.

Think of LSBot as having two settings. Fast Mode is its default: it’s quick, focused, and perfect for getting a simple answer fast. Deep Dive Mode is its opposite: it takes more time but gives you a more comprehensive answer.

Here’s a quick look at what sets them apart:

1. Speed: Fast Mode replies in seconds; Deep Dive may take 30+ seconds.
2. Scope: Fast Mode only references the page you’re currently on. Deep Dive Mode pulls from the entire Launch School curriculum and other resources like podcast transcripts and discussions.
3. Detail: Fast Mode provides quick, focused answers without any unnecessary extras. Deep Dive provides longer, more detailed responses, often including examples and deeper reasoning.

When should I use Deep Dive mode?

Here are three times Deep Dive is especially useful:

1. When Fast Mode falls short

Start in Fast Mode. It’s quick and gives focused answers, usually a paragraph or two. But if you want more detail — or if the answer feels incomplete — turn on Deep Dive.

2. When you want to explore

If you’re learning a challenging topic for the first time, Fast Mode helps you stay focused without getting lost in rabbit holes. But when you’re ready to ask why, how, or what’s the big picture, Deep Dive is the better choice.

3. When your question goes beyond the current page

If you need context from elsewhere in the curriculum, Deep Dive is essential. A few examples:

• You’re solving an exercise with nested loops and want to know where to review them.
• You’re reading about truthiness and wonder how it connects to short-circuiting operators, but short-circuiting operators are discussed elsewhere in the course.
• LSBot tells you it can’t answer your question.

As a general rule, start in Fast Mode and then switch to Deep Dive if you need more. If you always prefer Deep Dive, that’s fine too. This guidance is to help if you’re unsure.

A Peek Under the Hood

You might wonder why LSBot can’t give answers that are both fast and thorough. The short answer is that depth requires extra steps behind the scenes. Here are two of the biggest ones:

Query Classification

Not all questions are alike. Compare:

1. “How do I slice a string?”
2. “When should I apply for Capstone?”
3. “Review my code for this project.”

Even at a glance, you can see they’re very different. If a person were answering them, they’d need three different sets of instructions for how to approach each question type.

LSBot works the same way. In Deep Dive mode, the first step is to classify the type of question. Once we know what kind of query you’ve asked, we can feed LSBot detailed, tailored instructions — for example, following Launch School’s code review practices when it sees a request like #3, or focusing on examples at the right level in the curriculum when it’s a conceptual question.

This extra classification step produces sharper, more relevant answers — but it also takes time.

Retrieving Relevant Documents

In Fast Mode, LSBot only ‘sees’ the page you’re currently on, so it’s quick but limited. It can’t pull in context from other parts of the curriculum.

In Deep Dive Mode, it searches across the entire curriculum (and sometimes beyond) before answering, enabling richer, big-picture explanations at the cost of extra time.

We’re excited to see how students use this feature. Try it out and let us know what you think! We’re building LSBot to make your journey to mastery more engaging, efficient, and fun — and we hope it does that for you.

LSBot Updates: Deep Dive Mode, Question Boxes, and More was originally published in Launch School on Medium, where people are continuing the conversation by highlighting and responding to this story.

What is the Open Book Shelf?

The Open Book Shelf is a collection of over 15 publicly available books, most of which are assigned readings during the Core Curriculum. It gives prospective students a way to explore Launch School’s approach to learning before committing to the program. And now, thanks to LSBot, readers can go beyond passive reading and engage with the material in a more hands-on, personal way.

Why?

The Open Book Shelf is part of our effort to make high-quality learning materials freely accessible, while also helping students evaluate whether Launch School is the right fit for them. Over the past year, LSBot has become an important part of the student experience, supporting independent, mastery-based learning in nearly every part of the curriculum. It makes sense that prospective students should also be able to see — and experience — how LSBot fits into our pedagogy. To that end, we’ve integrated LSBot into these books in four key ways:

1. LSBot Questions

The first feature you’re likely to encounter in the books is LSBot Questions. These questions are designed to deepen your understanding and promote active learning. You can try answering them yourself and receive tailored feedback, or ask LSBot to walk you through the answer if you’re not quite ready to tackle them alone.

2. LSBot Conversations

LSBot Conversations — another favorite from the Core Curriculum — are now available in the books. These focused, quick exchanges are great for clarifying tricky ideas or checking your understanding as you go, without sending you down distracting rabbit holes. You can ask LSBot about anything in the chapter you’re working through.

3. LSBot Hints

At the end of most chapters in the Launch School books, you’ll find several exercises to help reinforce the chapter’s content. When you’re working through these exercises and get stuck, LSBot can offer helpful nudges without spoiling the answer, so you can keep making progress while still doing the hard thinking needed for true mastery.

4. LSBot Exercise Reviews

Finally, once you’ve completed an exercise, you can submit your work to LSBot for feedback. LSBot will help you evaluate whether your solution is correct and offer insights into possible improvements or alternative approaches.

These features are available immediately for free in every book on the Open Book Shelf. You don’t need to log in or register to start using them. We’re excited to share LSBot with more people, whether you’re new to Launch School or revisiting key concepts. Try it out and let us know what you think!

LSBot Joins the Open Book Shelf was originally published in Launch School on Medium, where people are continuing the conversation by highlighting and responding to this story.

LSBot Hints for Exercises

We recently introduced LSBot Conversations to help you work through assignments by asking targeted questions. Now, you can also get help on individual exercises, with one key difference: LSBot won’t give you the solution.

Struggling through an exercise is a core part of mastery-based learning. LSBot is now better equipped to help you stay on track, without doing the work for you. Use this new feature to get unstuck when you’re not sure how to begin, to pinpoint bugs in your code, or to explore ways to improve your solution.

Pop-out Modal for Better UX

Many of you shared that switching between LSBot and the assignment material was clunky. To address this, you can now pop LSBot conversations into a separate window, so you can reference the assignment while chatting with LSBot:

Note that this feature is not available when working on a smaller screen, such as a phone or small tablet.

LSBot Graded Quizzes

Written quizzes are now graded instantly by LSBot, giving you immediate feedback after submission. If you think LSBot misjudged an answer, you can leave a comment on the quiz, and a TA will review the flagged question.

Why the change? We recently expanded interviews at Launch School to offer more face-to-face time with TAs and better replicate real-world scenarios that you’ll encounter on the job hunt. This update continues in that direction, shifting TA focus toward higher-impact interactions — such as interviews and live study sessions — across a broader range of topics and times.

Quality-of-Life Updates

We’ve made a handful of small but meaningful improvements to make using LSBot feel smoother and more consistent.

• Better Conversation Continuity: LSBot has sometimes struggled with follow-up messages during conversations in assignments. We’ve rolled out a series of updates to improve this, making back-and-forth interactions more reliable.
• Improved Syntax Highlighting: Syntax highlighting for in-line code is now consistent and uses Launch School’s standard formatting as soon as your message is sent.
• A Range of UI Improvements: LSBot conversations now include a scroll bar, a larger default modal, and more consistent text sizing and formatting.

We hope these updates make your experience with LSBot smoother and more helpful. As always, your feedback drives these improvements, so don’t hesitate to share what’s working and what isn’t! As always, you can refer to the LSBot User Guide for detailed instructions and the most up-to-date information.

LSBot Updates - Exercise Hints, Instant Grading, and UX Improvements was originally published in Launch School on Medium, where people are continuing the conversation by highlighting and responding to this story.

We’ve just launched LSBot Conversations, letting you chat with LSBot as you work through assignments. These conversations are perfect for clarifying information about an assignment as you read through it. You can ask about specific passages, get a quick summary, or find out exactly where something is mentioned in a video.

If you’re familiar with LSBot on Slack, you’ll notice LSBot Conversations feel different — that’s by design. This article will highlight those differences, share tips for making the most of them, and explain when Slack might be a better fit. We’ll also introduce LSBot Questions, a companion feature intended to seed conversations and encourage active learning with LSBot Conversations.

To summarize how LSBot Conversations in the app differ from tagging or DMing LSBot in Slack, we can focus on three core traits: it’s fast, concise, and limited in scope. Let’s look at these characteristics and how they support the focused, in-the-moment learning we aim to support in the app.

# 1: Fast

Designing an LSBot feature is all about tradeoffs, and every interaction would be instant in a perfect world. With LSBot Conversations, we’ve chosen to prioritize speed, so you’ll typically get a response in 3–10 seconds, depending on your question. This emphasis on speed is meant to create a more conversational feel, helping you engage in quick, back-and-forth exchanges that clarify gaps in your understanding as they arise.

# 2: Concise

LSBot responses in the app will be much shorter than what you’re used to on Slack. This is partly about speed: by limiting the length of responses, LSBot can reply more quickly. When we imagine students using LSBot in the app, we picture someone reading an assignment for the first time, possibly feeling confused or overwhelmed. An extended, detailed response might feel more discouraging than helpful in that moment. Shorter answers allow LSBot to respond directly to your question without sending you down a rabbit hole. You can always ask follow-up questions to go deeper, or turn to LSBot on Slack if you’re looking for a more extensive explanation.

# 3: Limited in Scope

LSBot in the app won’t have as much context as LSBot in Slack. It will know the current assignment very well, but it won’t be able to make broad connections across assignments or courses in the curriculum. Why limit the scope? Keeping the focus narrow helps LSBot stay grounded in the current task, but it also helps us to accomplish those faster response times we discussed above.

Should I use an LSBot Conversation or Slack?

Let’s look at a few sample questions and discuss whether they are best asked in an assignment using an LSBot Conversation or in Slack:

Great questions for LSBot Conversations in the app:

Questions that would work in any assignment:

• Can you summarize this assignment? What are the key takeaways I should focus on?
• Can you create a few practice problems based on the material in this assignment?

Questions that work well when you’re on the relevant assignment:

• What does this sentence mean? “Operations involving operators with high precedence get evaluated before operations involving low precedence.”
• Why do we have to declare the counter variable outside of the increment function?
• At what point in the first “Watch Others Code” video does the student finish their PEDAC and start writing code?

Questions better suited for LSBot in Slack:

Questions that go beyond the current assignment:

• Where in the curriculum did we learn about closures before?
• What’s the assessment like for PY119?
• Will I use the concepts from LS170 in the front-end curriculum?

Questions about admin or non-curriculum questions:

• What happens if I receive a Not Yet?
• When can I apply for Capstone?
• How do I attend a SPOT session?

LSBot Question Boxes

You can initiate a conversation with LSBot from any assignment, anytime. In addition to using the “Ask LSBot” tab to start a conversation, there are also LSBot Questions sprinkled throughout the curriculum. These questions are meant to strengthen your mental models and encourage active learning. If you think you know the answer, try responding, and LSBot will provide feedback.

If you’re unsure or don’t care to answer, you can click on the question to send it to LSBot, which will answer the question for you.

We hope you’ll try out LSBot Conversations in the app to get quick, focused help while working through assignments. As always, you can find the most up-to-date information on LSBot in the LSBot User Guide.

Happy studying! 🚀

LSBot Conversations — Fast, Focused Answers in the Launch School App was originally published in Launch School on Medium, where people are continuing the conversation by highlighting and responding to this story.

We’re happy to announce the rollout of several new features that are available immediately for Launch School students using LSBot. First and foremost, LSBot is now available for direct messaging! The ability to use LSBot privately is undoubtedly the most anticipated update, but it’s not all we’ve been working on. Read on to hear about all of the new features out now:

• Direct messaging with LSBot
• Smart Prompts via Slack shortcuts
• Progress tracking for your queries
• Query Classification: A peek “under the hood”

In this article, we’ll dive into these new features and explore how they benefit students. Plus, we’ll give you a behind-the-scenes look at query classification and how it improves LSBot’s responses.

For concise instructions and images, you can always reference the LSBot User Guide.

LSBot in Your DMs

You’ve asked, and we’ve listened. While we love seeing the community fostered in our public LSBot channels, we understand that it can be intimidating to ask questions openly. With private messaging, you can now feel more comfortable using LSBot for anything — quick clarifications, in-depth debugging, or even questions you might think are too basic to ask in public.

Direct messaging with LSBot works similarly to posting in an #lsbot- channel, with a few key differences. Here’s what you need to do:

1. Search for LSBot as you would search for a fellow student to DM. You can also find LSBot underneath the list of “Apps” at the bottom of the left-hand navigation panel in Slack.
2. Send a message as you would in a public channel, except you no longer need to tag LSBot. LSBot will automatically respond to your message.
3. Continue the conversation in the thread. Just like in our public #lsbot- channels, LSBot will respond to your query in a thread. Start a new thread when you want to transition to a new topic or if LSBot has any unexpected or unhelpful behavior.

We hope you’ll give it a try and let us know how it goes!

Smart Prompts

This new feature helps you quickly format messages whose structure we see frequently. If you want your code reviewed for clarity, best practices, and errors, for example, you can call upon this new shortcut to have a form you can quickly fill out rather than typing out instructions for LSBot. This is also a great way to explore what LSBot can do, especially if you’re new to the scene.

To access Smart Prompts, you’ll use Slack’s “shortcuts” or “slash commands.” Both of these will allow you to access the same pop-up form. From any channel, type a forward slash (/), and a list of commands will appear. Select “Ask LSBot,” and a pop-up will prompt you to choose the type of question, the destination, and your message content.

The first time you use Smart Prompts, “Ask LSBot” may not appear in the list of shortcuts. If you don’t see it, start typing after the slash — entering “/ask” should bring up the shortcut you need.

You can also access Smart Prompts by clicking on the shortcut icon in the message box:

This is just a quick overview — refer to the LSBot User Guide for detailed instructions and demos.

Progress Tracker

It takes time to generate responses from LSBot, so we’ve added a progress tracker to give you insight into what’s happening behind the scenes. Now, when you send a message, you’ll see a step-by-step breakdown of how your query is being processed:

This small but useful addition gives you a clearer view of LSBot’s workflow. Additionally, you can now see how your query has been classified. If you’ve asked for a code review but your query has been classified as “Explain a concept,” you know that LSBot hasn’t quite gotten the point of what you’re asking. This can help you to refine your queries and give insight into when LSBot needs more guidance.

Query Classification

For those curious about LSBot’s inner workings, we’d like to introduce a new backend feature that’s making a big difference. While you won’t see query classification directly in the interface — aside from in the progress tracker, as showcased above — it plays a major role in improving the relevance and precision of LSBot’s responses.

To talk about query classification, let’s take a look at the steps that LSBot follows to process your queries:

1. You ask a question — LSBot receives your prompt (e.g., “What is variable shadowing?”).
2. LSBot searches for relevant Launch School content — LSBot will find the most relevant Launch School-specific information. This can come from the curriculum itself or other LS assets like podcast transcripts, Reddit threads, FAQs, and more.
3. Response Generation — The retrieved information, along with specific instructions, is passed along to generate a well-informed answer.

Now, let’s discuss where query classification fits into this picture. When using LSBot, most queries fit neatly into a handful of categories. A few that come to mind without writing an exhaustive list:

• Explaining concepts — “What is variable shadowing?”
• Code reviews — “Critique my solution…”
• Practice problem generation — “Create five problems…”
• Debugging help — “I’m getting this error when…”

Query classification means that we add a step between #1 and #2 where we use AI to “classify” your query into a specific category. The result of this classification is that every step in the process thereafter is improved. We can retrieve more relevant information, give more precise instructions for that type of task, and leave out irrelevant information. You can imagine how this improves the response quality!

As an example, imagine asking LSBot to review your code. With query classification, it will now prioritize checking syntax and best practices rather than retrieving theoretical explanations.

While this explanation only scratches the surface, it hopefully gives you a bit of insight into one of the many ways we’re working to improve LSBot and create a more impactful tool.

Enjoy the New Features!

We’re excited for you to explore LSBot’s new features. With every update, LSBot is becoming an even more powerful study tool. Our goal is to make your interactions more seamless, helpful, and personalized to your needs. As always, we look forward to hearing your feedback — let us know how LSBot is working for you and what we can improve next!

LSBot Now Supports Private Messages & More New Features was originally published in Launch School on Medium, where people are continuing the conversation by highlighting and responding to this story.

🤖 What is LSBot? 🤖

LSBot is an AI-based assistant you can tag on Slack and ask questions about Launch School and its curriculum, just like you might ask a TA.

It uses a large language model in combination with the Launch School curriculum to answer questions or help with other educational tasks. In addition to the Launch School curriculum, LSBot knows about other content like podcast episodes, interesting Reddit threads, past workshop recordings, interviews with Capstone grads, and more. LSBot also knows where you’re at in the curriculum. This means that when you ask a question, LSBot has enough context to give you a helpful answer. When you ask about solving a bug problem, LSbot won’t start prescribing neem oil or peppermint spray.

🎙️ Join a Live Info Session 🎙️

Want to see what LSBot can do? Join our upcoming live info session! You’ll see examples of LSBot in action, hear tips from students using it in their studies, and learn best practices for integrating it into your workflow. You’ll also have a chance to hear from Chris and ask questions about LSBot. Join us on Wednesday, December 11th at 6 pm Eastern.

Register Here.

🔎 Where Can I Use LSBot? 🔎

LSBot will be available in six Slack channels. Where you use LSBot will depend on the type of questions you want to ask:

Back-end Curriculum Questions

• #lsbot-javascript: For JavaScript-specific questions related to *back-end* courses. JS100-JS185 students should ask questions here.
• #lsbot-python: For Python-specific questions. PY100-PY185 students should ask questions here.
• #lsbot-ruby: For Ruby-specific questions. RB100-RB185 students should ask questions here.
• #lsbot-backend: For questions about courses that aren’t track-specific in the back-end curriculum, including LS170, LS171, LS180, and LS181.

Front-end Curriculum Questions

• #lsbot-frontend: For all front-end course questions. From LS202-TS249.

Non-technical Questions

• #lsbot-general: For any questions that aren’t related to a specific course. These could include things like study tips, Capstone questions, comparing tracks, and help with resources.

💻 How Do I Use LSBot? 💻

We put together a guide on using LSBot here! We recommend taking a look. It covers things like what LSbot can and can’t do, common pitfalls you can encounter when using LLMs, and sample templates you can use to get started.

In a nutshell, all you need to do is:

1. Navigate to the appropriate Slack channel where LSBot is available.
2. Tag LSBot by typing `@LSBot` followed by your question.
3. Continue the conversation by tagging LSBot in subsequent messages in the same thread.

🚧 A Note on Accuracy 🚧

Please note that LSBot is a work in progress. We’re working hard to make LSBot more reliable, accurate, and helpful, but it can make mistakes. LSBot is meant to be a supplemental tool, not a replacement for the Launch School curriculum. If you’d like to know more about the limitations of LSBot, check out the user guide.

We hope you enjoy LSBot and share your feedback with us. Happy studying!

LSBot Now Available For All Students was originally published in Launch School on Medium, where people are continuing the conversation by highlighting and responding to this story.