Vox

Leitura de 32 min

Using ChatGPT for Career Experience Simulations - LLMs and Direct Simulations for Faster Career Experience Gains

Summary

Experience is a critical teacher in career development, but waiting for real-life opportunities to practice important skills like critical conversations or technical interviews can take years. This episode explores how to accelerate experience gathering by creating low-stakes simulations, drawing inspiration from the aviation industry’s training methods where pilots practice emergencies in safe, simulated environments before facing real situations.

The host introduces two main approaches to simulation: direct simulations and LLM-based simulations. Direct simulations involve creating practice scenarios that closely mimic real situations, such as mock technical interviews with time constraints or presentation walkthroughs with trusted colleagues. These provide opportunities to iterate and learn in environments where failure has minimal consequences but offers significant learning benefits.

LLM-based simulations represent a newer approach where tools like ChatGPT or Claude can simulate conversations with difficult colleagues, bosses, or interviewers. By providing detailed prompts that describe personality traits and conversation goals, users can practice critical workplace conversations asynchronously. The host provides specific examples of prompts and explains how these simulations can be run multiple times, taken to extremes for better preparation, and used to identify weaknesses in communication or resume content.

The episode emphasizes that our brains process simulations and imagination similarly to real experiences, making simulated practice highly effective. By combining both direct and LLM-based simulations, professionals can create a rich tapestry of learning opportunities that accelerate skill development without waiting for rare real-world events. This approach provides asymmetrical benefits where the upside of learning far outweighs the minimal downside of simulated failures.

Practical applications include preparing for technical interviews, practicing difficult conversations with colleagues or managers, refining presentation skills, and identifying weaknesses in professional materials. The host recommends using direct simulations for qualitative feedback on body language and presentation style, while LLMs excel at content-focused practice and conversation simulation available anytime.

Recommendations

Methods

  • Direct simulation — Creating practice scenarios that closely mimic real situations, such as mock technical interviews with time constraints or presentation walkthroughs with trusted colleagues for feedback.
  • LLM-based simulation — Using language models to simulate difficult conversations by providing prompts that describe personality traits and conversation objectives, allowing for repeated practice in low-stakes environments.

Tools

  • ChatGPT/Claude — Large language models that can be used to simulate conversations with colleagues, bosses, or interviewers by providing detailed prompts about personality and conversation goals.
  • Wix Studio — A developer-first website builder with VS code-based IDE, GitHub integration, and AI code assistant for generating snippets and troubleshooting bugs.

Topic Timeline

  • 00:00:00 — The importance of experience in career development — The episode opens by discussing how experience is a critical teacher in career growth, particularly for rare events like critical conversations that might only occur once a month. The host explains that accumulating enough of these experiences to feel confident can take years, creating a bottleneck in professional development.
  • 00:01:52 — Aviation industry as a model for simulation training — The host draws parallels to aviation training, where pilots learn to handle emergencies through simulations rather than waiting for real-life failures. This creates a low-stakes environment where they can build muscle memory and confidence without actual danger, demonstrating how simulation accelerates experience gathering.
  • 00:06:24 — Introduction to creating low-stakes environments for career growth — After the sponsor segment, the host introduces the core concept: creating low-stakes environments to accelerate experience gathering in software engineering and leadership careers. The discussion focuses on how simulations provide asymmetrical benefits where learning upside far outweighs the downside of simulated failures.
  • 00:11:41 — Two tactical approaches: direct simulations and LLM-based simulations — The host outlines two main methods for simulation. Direct simulations involve practice scenarios that closely mimic real situations, like mock interviews with time constraints or presentation walkthroughs with colleagues. These provide iterative learning opportunities in safe environments where failure has minimal consequences.
  • 00:15:52 — Using LLMs like ChatGPT for conversation simulation — The host introduces LLM-based simulations as a newer approach, explaining how tools like ChatGPT can simulate conversations with difficult colleagues, bosses, or interviewers. By providing detailed prompts describing personality traits and conversation goals, users can practice critical workplace conversations asynchronously and repeatedly.
  • 00:18:48 — Practical exercise for LLM simulation with difficult conversations — A specific exercise is presented: think of a person you need to have a critical conversation with, describe their personality and the conversation goal to an LLM, and practice the interaction. The host provides example prompts and explains how this can reveal communication patterns and areas for improvement in a low-risk environment.
  • 00:22:26 — Taking simulations to extremes for better preparation — The host discusses how simulations can be designed with exaggerated characteristics to build greater resilience. If you can handle an extreme version of a difficult personality or situation, you’ll be better prepared for the more realistic version. This approach helps identify weaknesses and build confidence through challenging practice scenarios.
  • 00:25:50 — Comparing direct and LLM simulations: when to use each — The host compares the two approaches, noting that LLM simulations are always available and don’t require another person to act, while direct simulations are better for qualitative feedback on body language, warmth, and presentation style. A combination of both methods creates the richest learning opportunities for career development.

Episode Info

  • Podcast: Developer Tea
  • Author: Jonathan Cutrell
  • Category: Technology Business Careers Society & Culture
  • Published: 2024-12-03T08:00:00Z
  • Duration: 00:31:24

References

Podcast Info

Transcript

[00:00:00] Experience is an excellent teacher.

[00:00:15] You know this, of course, this is one of the reasons why years of experience is

[00:00:21] a very typical requirement on any job posting and rightfully so.

[00:00:27] Experience is hard to replace with any kind of academic learning, for example.

[00:00:34] That’s not to say that academic learning is not useful.

[00:00:38] It’s not to say that you can’t gain experience through other

[00:00:42] means other than time, but experience is a critical part of your career.

[00:00:50] One of the reasons that time and experience over time is one of the

[00:00:56] critical factors for success in your career is because there aren’t a ton of

[00:01:02] opportunities in your kind of day-to-day work to have, for example, critical

[00:01:08] conversations.

[00:01:10] You may have a critical conversation once a month, for example, and so over

[00:01:17] time, you may need to accumulate, let’s say 20 or 30 of those critical

[00:01:22] conversations before you can confidently approach any given critical conversation.

[00:01:29] This kind of waiting around is difficult to do.

[00:01:34] It takes a long time to get to the place where you’ve had enough of these kinds

[00:01:41] of discussions to feel confident.

[00:01:44] I want to switch gears and talk for a moment about a completely different

[00:01:48] industry, and you’ll see why in a few minutes.

[00:01:52] The airline industry or even just the aviation industry, flying private

[00:01:58] airplanes.

[00:01:59] If you’ve been listening to this podcast for a while, you know that I have my

[00:02:03] private pilot’s license and I’ve been flying for most of my life, either flying

[00:02:08] myself or flying as a passenger with my father, who also has his pilot’s

[00:02:13] license and has flown many different aircraft.

[00:02:16] So by nature of this kind of exposure, I am familiar with this industry.

[00:02:20] And you might know that aviation, especially the airline industry is

[00:02:26] known for its safety record.

[00:02:28] Some might disagree that this is true for all of aviation.

[00:02:33] Certainly it changes as you go through less training, but the training is the

[00:02:38] part that I want to talk about.

[00:02:41] So in order to operate an airline safely, in order to operate a large, you

[00:02:48] know, 200 passenger jet, for example, you need to know how the jet behaves.

[00:02:54] But not just when the jet is operating correctly, not only when the jet is, you

[00:03:02] know, doing what it’s expected to do, all systems operational, you know, no red

[00:03:07] lights blinking on the dash, so to speak.

[00:03:10] You also need to know what happens when something fails.

[00:03:14] How should I respond to it?

[00:03:16] How do I deal with multiple failures?

[00:03:19] What if I lose an engine or what if my hydraulic system fails?

[00:03:23] Now we don’t have to enumerate all the possible failures of a, you know, 737

[00:03:28] on this show, but the important thing to understand is these pilots know how to

[00:03:34] deal with these emergencies.

[00:03:36] And as it turns out, these pilots know how to deal with these emergencies

[00:03:40] through a type of experience.

[00:03:42] But very rarely is a pilot experiencing actual real life emergencies multiple

[00:03:49] times in their career, certainly not enough to learn through direct experience.

[00:03:55] So what is it that they are doing to get them this experience, to get them ready

[00:04:01] for dealing with an emergency?

[00:04:03] I’m sure this comes as no surprise.

[00:04:05] This happens as a part of their training, as a part of their learning process.

[00:04:11] And it happens most importantly, in a low stakes environment.

[00:04:17] This is something that very often we entirely ignore in our careers

[00:04:23] as software engineering leaders.

[00:04:26] Today, I’m going to teach you how to think about creating a low stakes

[00:04:31] environment to accelerate your experience gathering right after we

[00:04:36] talk about today’s sponsor.

[00:04:41] I’m going to ask you to stay put for five seconds longer than you would when I

[00:04:47] say the words website builder.

[00:04:50] I know what you’re thinking.

[00:04:52] It probably has limited control or outputs, a bunch of dirty code.

[00:04:57] This is my nightmares from the dream weaver days, but what about a node

[00:05:03] JS based builder that lets you add full stack JavaScript code to each of

[00:05:08] site with WIC studio, you can spend less time on your UI, uh, less time on

[00:05:18] hosting, less time on security, more time on the custom logic and the

[00:05:21] functionality that actually matters.

[00:05:24] That makes a difference to you and your client.

[00:05:28] You can develop in your preferred coding environment, whether that’s online

[00:05:32] in a browser and a VS code based ID or locally through a GitHub integration

[00:05:37] using whatever tools you’re used to either way with WIC studio, you’re

[00:05:41] deploying in a click extend and replace hundreds of powerful business solutions

[00:05:47] and custom built features with APIs and integrations.

[00:05:51] And when you need to speed things up, WIC studios, AI assistant is on hand

[00:05:55] to generate tailored code snippets, troubleshoot bugs, and retrieve product

[00:05:59] answers in just a few seconds.

[00:06:02] All of that is neatly wrapped up in an automatically maintained

[00:06:05] infrastructure for total peace of mind, working in developer first ecosystem.

[00:06:10] Head over to WIC studio.com.

[00:06:12] That’s W I X studio.com.

[00:06:15] Thanks again to WICs for sponsoring today’s episode.

[00:06:24] So we’re talking about, uh, creating a low stakes environment and we mentioned

[00:06:29] the aviation industry as a good model or, or a kind of a mental model for this.

[00:06:34] And the way that they do this in their training is they create simulations.

[00:06:40] They, uh, build the situation, whether that’s in the air or on the ground.

[00:06:46] For example, a simulated engine failure.

[00:06:50] This might just mean pulling your throttle, uh, you know, pulling the

[00:06:54] power on the engine all the way back to idle.

[00:06:56] This is essentially the same as losing the power to your engine in real life.

[00:07:02] But it maintains a level of safety because your engine is not actually turned off.

[00:07:09] This is important because you can go through the same motions that you would,

[00:07:12] and the airplane is going to respond in the same ways as it would if

[00:07:17] your engine actually failed.

[00:07:19] The simulation allows you to learn and create the muscle memory that you would

[00:07:24] need in order to actually respond to a real life engine failure.

[00:07:30] So I want you to think about this specifically in terms of simulation,

[00:07:35] right?

[00:07:35] Of course, also, especially in the airline industry, it’s a lot more cost

[00:07:40] effective, even if you could simulate these things, uh, in real life up in the

[00:07:46] air, uh, if you could simulate them flying at 7 37, this would be very

[00:07:52] cost prohibitive to do, right?

[00:07:54] Uh, it’s very unlikely that you’re going to do this kind of simulation

[00:07:58] with passengers on the airplane, for example.

[00:08:01] And so the airline industry relies on actual simulators.

[00:08:06] These are, uh, you know, these very complicated devices, uh, you know, that,

[00:08:11] that are the size of a, of a room, uh, that you would sit in and it has all of

[00:08:16] the, uh, high fidelity kind of instruments that you would expect to see in the real

[00:08:21] airplane, right?

[00:08:23] So the simulation has been created to give you a high level of immersion,

[00:08:30] right?

[00:08:30] This, this concept of immersion is that your brain is going to, uh, you know,

[00:08:34] operate in the simulation in a very similar manner, right?

[00:08:39] You’re going to build some connections in your brain.

[00:08:42] And here’s the important part.

[00:08:43] Okay.

[00:08:44] You might think, okay, well, that’s all fake, but, uh, when it comes to real

[00:08:47] life, how do we know that that’s going to transfer?

[00:08:50] There’s actually very good research.

[00:08:53] Okay.

[00:08:53] Very good research on simulation.

[00:08:55] And as it turns out, this kind of overlaps with research on imagination.

[00:09:01] Our brains are not very good at determining what is real, quote, real,

[00:09:08] and what is not real, what is either imagined or simulated, even if you can

[00:09:14] cognitively recognize, right?

[00:09:17] Like you, you can reason, you can rationally think through, okay,

[00:09:21] this is a simulation or this is, uh, my imagination, your brain will still

[00:09:27] decode your imagination and simulations as something very real in a very kind

[00:09:34] of related sense, uh, we, we experience high levels of emotion when hearing

[00:09:41] stories, sometimes these stories are not real at all.

[00:09:44] You can imagine there’s somebody getting emotional, watching a

[00:09:47] movie or reading a book, right?

[00:09:48] We can connect to, uh, something that is entirely imagined, entirely, you

[00:09:54] know, built up in our head and, uh, we, we connect to that.

[00:09:57] And so, uh, our brains experience that as, as real as anything else.

[00:10:03] Similarly, we can imagine the future and, uh, get excited about the future.

[00:10:09] We haven’t experienced it yet, uh, but we anticipate and that anticipation

[00:10:15] gives us the same kinds of feelings that we would have if we were actually

[00:10:19] experiencing the thing, right?

[00:10:21] The, the other side of the coin is, uh, we can anticipate and feel anxiety.

[00:10:26] And we can feel, you know, our, our hands get clammy, uh, our heart rate

[00:10:30] increases when we’re just thinking about an uncomfortable situation or, you

[00:10:36] know, a meeting that’s coming up that we’re not really looking forward to.

[00:10:39] Maybe it’s going to be a little bit of a stressful meeting.

[00:10:41] We imagine that meeting, uh, even though it’s not happening to us, we could be

[00:10:46] in a totally safe and secure environment.

[00:10:48] And when we start imagining it, our body starts to produce the fight or flight

[00:10:53] response that happens when we’re in a threatening situation.

[00:10:57] And so your brain is decoding your imagination as a threat, right?

[00:11:02] So, uh, there’s, there’s plenty of evidence that shows that this is the

[00:11:05] case that imagination simulation are both, uh, you know, very difficult to,

[00:11:11] um, you know, to, to parse through for your brain at its lowest levels.

[00:11:15] So we can take advantage of this reality, right?

[00:11:18] We can take advantage of it and we can simulate these kinds of situations,

[00:11:23] these critical conversations, for example, right?

[00:11:26] We can simulate these situations that we face in our work.

[00:11:30] Uh, interviews is another great example, right?

[00:11:33] So I’m going to share with you two different tactical ways that you can

[00:11:37] simulate these, these kinds of things in your daily work.

[00:11:41] The first is a kind of direct simulation.

[00:11:44] Okay.

[00:11:45] Direct simulation.

[00:11:45] What this looks like is let’s imagine that you are preparing for technical

[00:11:49] interviews, you may give yourself a leak code problem.

[00:11:53] This is something you’re probably already doing.

[00:11:55] You may not realize that you are doing a simulation when you do this.

[00:11:59] Uh, if you don’t feel like it’s a simulation, one thing you can do is you

[00:12:02] can constrain yourself on the amount of time that you take, right?

[00:12:06] So, uh, instead of just giving yourself, you know, an endless amount of time,

[00:12:10] imagine that there are people watching.

[00:12:12] And in fact, you can even have somebody, uh, watch you, you can, uh,

[00:12:17] contact a close colleague, maybe somebody in the developer tea community,

[00:12:22] uh, on our discord, you can contact somebody there and ask them if they

[00:12:26] would be willing to do kind of a mock, uh, technical interview with you.

[00:12:31] All right.

[00:12:31] So, uh, people are watching you.

[00:12:33] So you get a little bit of that feeling of somebody is judging the work that

[00:12:36] you’re doing, uh, rather than, you know, treating this just as, oh, I’m

[00:12:40] learning the material, you could also prepare for the other side of this

[00:12:45] interview, which is kind of the pressure, the feeling of pressure coming

[00:12:49] from another person watching.

[00:12:50] So try to emulate, right?

[00:12:54] Notice that we’re kind of mixing a little bit of, of terminology here,

[00:12:58] but emulation is trying to make the, uh, the situation that you’re in look a

[00:13:03] lot like the other, you know, the real situation they’re trying to simulate.

[00:13:07] Right.

[00:13:08] So try to emulate some of those properties, uh, emulate the amount of

[00:13:12] time that it might take emulate, you know, the number of people or, you know,

[00:13:16] the kinds of questions that those people are going to ask you try to create,

[00:13:20] uh, uh, you know, a mock version of the same thing.

[00:13:23] This is a direct simulation.

[00:13:25] Another example of a direct simulation might be like a mock

[00:13:29] presentation walkthrough, right?

[00:13:31] You have a press and a presentation to give to your colleagues.

[00:13:36] Uh, you have a colleague that you trust that you want to give this

[00:13:40] presentation to in advance and have them provide you critiques, have them take

[00:13:44] notes during the presentation, identify places that feel confusing.

[00:13:49] These are ways that you can gather feedback and learn, right?

[00:13:52] That’s the important part here, right?

[00:13:55] The, the idea is that you’re going to get multiple cycles of experience out of a

[00:14:00] single, you know, critical event, in this case, a presentation or, uh, an interview.

[00:14:06] And you’re going to get this experience by iterating through lower stakes

[00:14:10] environments.

[00:14:10] So what happens if you completely fail your leak code, uh, fake interview?

[00:14:17] This is the low stakes part of this learning, right?

[00:14:21] If you fail your leak code interview, the only thing that you really lose from

[00:14:26] that is a, uh, a feeling of pride, maybe, right?

[00:14:30] Or, or impressing, um, you know, this, the stranger that is doing the mock

[00:14:35] interview, right?

[00:14:36] There’s very little that you lose.

[00:14:40] Uh, but you gain the learning, right?

[00:14:43] This is why this is kind of an asymmetrical benefit to you.

[00:14:47] Okay.

[00:14:48] The asymmetrical side of this is you have a lot more to gain than you have to

[00:14:54] lose in a simulated environment.

[00:14:57] The gaining, uh, you know, the upside of, of a simulation is much larger than the

[00:15:02] downside and a failure state in the simulation.

[00:15:06] So these are, these are direct simulations.

[00:15:08] These are the ones that are more intuitive.

[00:15:09] These are the things that you would expect to do in order to simulate something,

[00:15:13] you know, uh, getting dressed in your interview attire and actually sitting

[00:15:18] in front of a zoom camera, having somebody interview you in a mock interview.

[00:15:22] These are ways that you could simulate your, uh, you know, that, that critical

[00:15:27] event, another example, right?

[00:15:29] And this is one that is a little less intuitive.

[00:15:33] Okay.

[00:15:33] That a little less intuitive.

[00:15:35] The previous ones were intuitive because basically you’re just trying to do the

[00:15:39] thing before you do the thing, right?

[00:15:41] You’re, you’re trying to create an opportunity to practice.

[00:15:46] These are, you know, the most basic kinds of practicing this second category is

[00:15:52] fairly new because you’re going to use something like chat GPT or Claude or

[00:15:58] whatever LLM of your choice.

[00:16:01] Okay.

[00:16:02] Um, preferably one of the newer models, the, the models have gotten to the place,

[00:16:07] especially more recently where this can be done to some good effect.

[00:16:12] Okay.

[00:16:12] And I hesitate to talk about specific tools on this podcast because it’s very

[00:16:17] possible that this goes out of date or, you know, I don’t want to, uh, produce

[00:16:22] clickbait for the show, but I do think that this tool set, uh, it represents a

[00:16:28] critical change in the way that we can simulate, uh, our environment.

[00:16:33] And here’s the reason.

[00:16:34] Okay.

[00:16:35] LLMs by their kind of fundamental nature.

[00:16:38] We’re not going to get into it and break down all of the

[00:16:40] specifics of how LLMs work.

[00:16:42] Uh, you can find that content elsewhere, but they are, uh, particularly

[00:16:47] good at understanding language.

[00:16:51] Okay.

[00:16:52] They are particularly good at understanding how humans behave, how

[00:16:57] they talk, how language evolves over this course of a conversation.

[00:17:02] Now understanding that, you know, an LLM can do this kind of language, uh,

[00:17:09] prediction, or it can understand a conversation.

[00:17:13] It can understand semantic meaning and as it moves through a conversation,

[00:17:18] it understands how the conversation is changing.

[00:17:21] Right.

[00:17:22] If we imagine, okay, I want to have a simulated conversation, LLMs provide

[00:17:29] a pretty amazing opportunity to simulate conversations.

[00:17:34] In other words, what we have in LLMs is something that is particularly good

[00:17:38] at understanding how a conversation may evolve.

[00:17:43] All right.

[00:17:44] You provide your responses, you provide some kind of, uh, you know, written

[00:17:49] input and the LLM is going to use a ton of information about how other people

[00:17:57] have responded to similar words.

[00:18:00] Okay.

[00:18:01] Similar construction, similar kind of logical construction.

[00:18:05] It’s going to use that to provide a response to you.

[00:18:09] In other words, in many ways, an LLM is fundamentally a conversation simulator.

[00:18:15] So here’s an exercise I want you to try.

[00:18:17] Okay.

[00:18:17] And everybody can benefit from this, not just people who are looking to, you

[00:18:20] know, uh, improve their interviewing skills.

[00:18:22] Okay.

[00:18:24] You are, or you’re especially going to benefit from this.

[00:18:27] If you have an asynchronous, uh, communication environment in your work.

[00:18:31] And most people who are listening to this episode, you do have this somewhere.

[00:18:36] Okay.

[00:18:36] Somewhere, uh, in your work, you have an asynchronous, uh, you know, work

[00:18:41] environment, or you have an asynchronous, uh, communication

[00:18:44] environment as a part of your work.

[00:18:46] All right.

[00:18:47] So here’s the exercise.

[00:18:48] I want you to think about a person that you need to have a critical conversation

[00:18:54] with, or that you don’t really get along very well with at work, but you have a

[00:19:00] incentive, you have some kind of incentive to get along with them or to,

[00:19:05] uh, persuade them or to talk to them about some particular topic.

[00:19:10] Maybe you need to provide them with a status update.

[00:19:14] Uh, you know, maybe, maybe this diamond dynamic doesn’t have to be your boss,

[00:19:18] right?

[00:19:18] And you don’t have to simulate just conversations with your boss.

[00:19:22] You could simulate a conversation with a coworker.

[00:19:24] You can simulate a conversation with somebody in a different department,

[00:19:28] right?

[00:19:28] Whatever it is, your situation, try to understand, uh, or, or, or nail

[00:19:34] down who that person is.

[00:19:36] Okay.

[00:19:36] What is it that you want to get out of this conversation with them?

[00:19:40] And why is it that it hasn’t worked before?

[00:19:43] What about their personality?

[00:19:45] Is it that’s difficult for you?

[00:19:47] All right.

[00:19:48] So I want you to try to describe that person as your prompt for chat

[00:19:54] GPT or Claude or whatever LLM you’re using.

[00:19:58] An example prompt might look something like this.

[00:20:02] You are a hard-nosed deadline focused boss, and we’re going to have a

[00:20:09] conversation about, uh, you know, a body of work, a project that I’m trying

[00:20:16] to deliver that is behind schedule.

[00:20:19] It’s behind schedule by a month or two and you want hard deadlines.

[00:20:26] You want to push me for that.

[00:20:28] You’re going to corner me and force me to provide these dates.

[00:20:34] And I want you to do this as if we are sending each other Slack messages.

[00:20:40] Okay.

[00:20:40] This kind of prompt, uh, gives the LLM a kind of a starting, a seed personality.

[00:20:48] And you will likely be surprised, um, at how well this simulates a conversation.

[00:20:54] So you start the conversation, uh, and you ask them to kind of ask

[00:20:59] you for an update, for example, right?

[00:21:01] Now what’s really interesting is that the LLM can also, uh, you know, you

[00:21:07] know, you could, you can run the simulation multiple times.

[00:21:09] You can carry the conversation to its natural endpoint and then ask the LLM.

[00:21:14] Okay.

[00:21:15] Let’s start over.

[00:21:16] You may also be able to ask the LLM something like, help me understand

[00:21:22] where I’m going wrong or help me understand what part of the conversation

[00:21:28] I need to change my approach on.

[00:21:31] Now, this isn’t a perfect system, of course, just like almost every AI

[00:21:35] product you’re using right now.

[00:21:37] There’s a big old disclaimer on this episode, uh, that basically, you

[00:21:41] know, it, it might work well or it might not.

[00:21:44] You, you can certainly trick this prompt, uh, you know, to, to tell

[00:21:48] you something that you want to hear.

[00:21:50] So I want you to exercise caution with this.

[00:21:53] But if you try to treat this as a true simulation, in other words, imagine

[00:21:57] that you are really actually messaging that person, right?

[00:22:01] That person that you thought of, uh, that, that you have this kind of conflict

[00:22:06] with, or you need to have a critical conversation with, if you actually

[00:22:09] truly imagine yourself talking to them in this way, you’re likely to

[00:22:14] gain some kind of benefit.

[00:22:15] You’re likely to gain some kind of insight that you didn’t have before.

[00:22:21] Now, a really important part of simulation is that you can take things to the extreme.

[00:22:26] And this kind of dovetails into our previous discussions about different coaching

[00:22:31] voices or those hats that I was telling you to wear.

[00:22:33] Remember we had the hyper-optimizer, uh, you know, coaching voice that allowed

[00:22:39] you to think through the lens of somebody who is only, who only cares

[00:22:42] about a specific outcome, right?

[00:22:44] They only care about, you know, getting more money.

[00:22:46] For example, you can design in this LLM situation, right?

[00:22:52] You can design, uh, this person to be a more caricaturized or a more extreme

[00:22:59] version of the things that you’re, you’re struggling with in real life.

[00:23:04] Why would you want to do this?

[00:23:06] Well, as a general rule, if you are able to deal with something more extreme,

[00:23:12] then you are also going to be able to deal with something less extreme.

[00:23:16] If you can mentally prepare yourself for somebody who exhibits this particular

[00:23:22] characteristic in a very extreme manner, then you can also prepare for the more

[00:23:28] realistic version that you’re more likely to encounter in real life.

[00:23:33] So in my previous example, a boss that only cares about deadlines, very

[00:23:39] rarely is this actually true.

[00:23:41] There are, you know, most of the time somebody who’s asking you for a

[00:23:46] specific date of delivery, they have multiple things that they really care

[00:23:51] about, but they’re asking you for a date because that’s what they’re being asked

[00:23:55] for, right?

[00:23:56] So it makes sense to simulate both.

[00:23:58] It makes sense to simulate both the extreme caricaturized version, uh, of,

[00:24:04] you know, whatever this difficult personality is that you’re dealing with.

[00:24:08] Uh, or, or maybe, you know, you’re, maybe you want to do some kind of written

[00:24:11] essay, uh, simulations, you know, tell telling the LLM to ask you certain

[00:24:17] questions about your background.

[00:24:20] Uh, another example, if you’re preparing for interviews, feed your, uh, your

[00:24:25] resume to chat GBT or to, to Claude and ask it to, you know, be a pessimist and

[00:24:33] to poke holes, to find places, uh, that they can ask you questions to make you

[00:24:39] feel uncomfortable or make you feel, uh, you know, inferior.

[00:24:43] Now this, this is, this sounds like it would be really difficult, but again,

[00:24:47] here’s the important thing.

[00:24:48] This is a low stakes environment, right?

[00:24:51] This is a simulation is a low stakes environment.

[00:24:53] There’s nothing that’s actually bad that will happen as a result of this chat.

[00:25:00] Right.

[00:25:01] Very likely.

[00:25:02] The thing that it will do is it will point out areas that you can improve on.

[00:25:07] It’s going to point out, you know, patterns of, of communication that

[00:25:11] maybe you didn’t realize you had, or it might point out, oh, this is a flaw or

[00:25:15] this is a thin part of my resume, or this is a confusing wording that I’ve put in

[00:25:20] here, or I’m being vague about, you know, this particular job or, or the outcomes

[00:25:25] that I’ve produced in this role.

[00:25:27] So there are a lot of ways that you can use this.

[00:25:30] And again, I have this big disclaimer because, um, you know, these, these things

[00:25:34] are not really deterministic and you should be using them with a little bit

[00:25:39] of caution, but with the intent of trying to find ways to simulate your real life

[00:25:46] situations in a much lower stakes environment, right?

[00:25:50] So you might be asking, okay, what, why would I use that versus the direct

[00:25:54] simulation one it’s available all the time, right?

[00:25:57] You can, you can use this at midnight.

[00:26:00] Uh, you know, the night before an important meeting, if you really needed

[00:26:04] to do that, uh, it’s available to you, right?

[00:26:07] Uh, another thing is that if you ask another person to act like, uh, you

[00:26:13] know, uh, a particular person.

[00:26:16] So if you’re asking a colleague or, uh, you know, a coworker or maybe somebody

[00:26:21] outside of your company, if you’re asking them to act like a boss, they may not be

[00:26:26] able to, uh, to act very well, right?

[00:26:29] So it kind of breaks a little bit of that simulation.

[00:26:31] So with an LLM, the acting portion is, you know, there’s, there’s not like a, an

[00:26:38] association that you have to get past with an LLM, so there’s no kind of

[00:26:42] personal connections that muddy the waters with this.

[00:26:46] Now you may want to use a direct simulation when you’re trying to

[00:26:49] get more qualitative feedback, right?

[00:26:52] Um, you know, this, and when I say qualitative, I mean, feedback about,

[00:26:57] for example, your, your body language.

[00:26:58] This is something that LLMs are not really trained well to do.

[00:27:02] Uh, if it’s, if it’s dependent on something more than just your pure

[00:27:06] language that you’re using, then a direct simulation may be a better choice.

[00:27:11] So, uh, as an example, if you’re simulating, um, in your interview

[00:27:16] process, you may be able to use an LLM simulation to find content problems.

[00:27:23] Right.

[00:27:24] If you want to tell a story or tell about your background, the LLM may be able

[00:27:28] to help you shape your content, but your actual presentation, the warmth, uh, that

[00:27:35] you present or the confidence that you show up with your posture, all of these

[00:27:39] other kind of more qualitative things are going to be better

[00:27:43] assessed by an actual person.

[00:27:46] Overall, a mix of these tools is going to provide you with a much richer

[00:27:52] tapestry of simulation opportunity.

[00:27:57] Right.

[00:27:57] So this idea that you have to wait for something to happen in your career, you

[00:28:01] can dispose of that idea and instead you can seek out an opportunity to learn

[00:28:06] about it through a simulated situation.

[00:28:09] This is a much lower stakes, much lower cost and a much higher iteration rate.

[00:28:14] Thank you so much for listening to today’s episode.

[00:28:17] I hope you enjoyed this discussion about simulation.

[00:28:19] Thank you again to today’s sponsor Wix Studio.

[00:28:23] Website builders of yesterday are gone.

[00:28:26] With Wix Studio’s developer-first ecosystem, you can spend less time on

[00:28:30] tedious tasks and more on the functionalities that matter the most.

[00:28:35] Develop online via a VS code based IDE in the browser, or you can develop

[00:28:41] locally using your tool set and GitHub.

[00:28:44] Extend and replace a suite of powerful business solutions and ship faster

[00:28:48] with Wix Studio’s AI code assistant.

[00:28:51] All of that is wrapped up and maintained in an automatic infrastructure

[00:28:55] for total peace of mind.

[00:28:57] Work with developer-first ecosystem at wixstudio.com.

[00:29:00] I want to make a quick three part ask of you at the end of the year here.

[00:29:05] And hopefully this episode and other episodes of developer T have made a

[00:29:09] difference in your life that these three things, which are totally free for you

[00:29:12] to do are doable, uh, on, on your end.

[00:29:16] Um, and these are the ways that we keep the show alive.

[00:29:18] The first thing is help another engineer by sending them the show.

[00:29:24] If you know of an episode that might help them, right?

[00:29:27] Not, don’t just send them the show and say, you should listen to this.

[00:29:30] Help them out.

[00:29:31] Find an episode that makes sense.

[00:29:33] Maybe you know somebody who’s going through interviewing process.

[00:29:36] This episode might be useful to them.

[00:29:38] Right.

[00:29:39] So send an episode to a person that you know, that might benefit from it.

[00:29:43] It doesn’t have to be an engineer.

[00:29:45] Uh, sometimes these concepts easily transfer to other, uh, to other disciplines.

[00:29:50] Certainly, uh, this episode will, will transfer to other disciplines.

[00:29:55] That’s, uh, that’s the first thing.

[00:29:56] The second thing is leave a review in iTunes.

[00:29:59] We’ve stopped asking you to leave a review in every platform on the planet.

[00:30:03] That’s because iTunes is the vast majority, uh, of listeners for this show.

[00:30:08] And if you leave a review anywhere, iTunes is going to have the, the largest impact.

[00:30:13] Uh, so leave a review in iTunes.

[00:30:15] You can of course still have a review in those other places, but, uh, asking you

[00:30:19] to leave a review in iTunes, this is the most important thing, uh, to help other

[00:30:24] engineers find this show and decide to listen to it.

[00:30:27] The last thing I want you to do is to find a day between now and the end of

[00:30:32] the year to spend 10 to 15 minutes disconnecting and thinking back over the

[00:30:40] year, and, uh, you can use this time also to think forward, but really reflecting

[00:30:46] on how your year has aligned with your priorities in your career.

[00:30:53] And here’s the thing.

[00:30:54] Developer T directly doesn’t benefit from you doing this.

[00:30:57] You benefit from this.

[00:30:59] And I hope that as you move forward in your career, and as, especially if you’ve

[00:31:04] listened to the show in the past, that you’ll continue listening, that this show

[00:31:08] has been a part of the positive side of your year, uh, since last December.

[00:31:14] Thanks so much for listening and until next time, enjoy your tea.

Visão de gráfico