我是初级开发者，资深开发者告诉我，AI 将在 3 年内取代初级编程工作。

顾问们就初级开发岗位是将会消失还是仅仅发生变化展开了辩论。尽管他们在时间线和严重程度上的看法不一，但他们达成了一项关键共识：初级岗位的角色不会消失，而是正在转变为某种不同的形式——审查 AI 输出、调试生成的代码以及理解复杂系统。真正的风险并非失业，而是除非你属于能够驾驭 AI 集成工作流的上层梯队，否则你的收入可能会显著低于几年前初级开发人员的水平。

The Contrarian WARN
Three years is fantasy. I lost everything in 2008 listening to senior people tell me what was 'definitely' about to happen. Here's what they're not seeing: junior jobs aren't disappearing, they're mutating. The evidence shows routine coding gets automated, but someone still has to understand the system, catch the AI's mistakes, and know why the generated code is garbage. That's still a junior's job, just different.

Marcus Sterling REFRAME
Everyone's fixating on whether jobs vanish or not, but the real damage is already happening—wage suppression. IBM tripled their Gen Z hiring, sure, but nobody's asking what they're paying those kids. When AI lets one senior dev do the work of three juniors, companies don't eliminate the entry-level roles, they just slash what they're willing to pay for them. The person asking this question might get a job, but they'll be making 60% of what juniors made five years ago while being expected to review AI output at senior-level speed.

Yuki Shin PUSHBACK
Marcus, you're half right about wage suppression but you're missing what I'm actually seeing in my budget cycles. The Morgan Stanley report shows enterprises are building more complex applications because of AI tools, which means I need juniors who can handle that complexity—not just cheaper labor. I'm paying the same or more for juniors who can debug AI output and integrate it into production systems, because that skill is rarer than you think.

Dr. Aris Thorne-Flux PUSHBACK
Yuki, your budget cycles are lagging indicators—you're describing what you're paying now, not what the market will bear in eighteen months. I've watched this exact pattern in three prior automation waves: first the roles stay, then compensation fragments into tiers where only the top quartile maintain wages while everyone else craters. The evidence shows junior developer roles dropped 46% already, and that's before most enterprises finish integrating AI workflows into hiring pipelines.

顾问们就 AI 工具是否抑制初级开发者薪资，或为能有效使用它们的人创造优质职位展开了辩论。Contrarian 和 Yuki 认为，公司正在为具备 AI 技能的初级开发者支付更高薪酬，因为他们能带来更高的生产力；而 Marcus 和 The Auditor 则反驳称，初级人员整体招聘量下降了 46%，公司正从每位员工处获取更多产出，同时总成本更低。核心分歧在于：少数被录用的初级开发者薪资上涨，是否足以抵消可用职位的大幅减少。

The Contrarian WARN
Aris, compensation fragmentation already happened — in 2001, after the dot-com crash. Juniors who survived that made less for years, sure, but the ones who couldn't were gone in six months. The real tell isn't wages dropping across the board, it's that they've already stopped hiring juniors who can't work with AI tools at all. I'm watching postings that used to say 'entry level' now require 'experience debugging LLM output' — that's not a lagging indicator, that's a filter that didn't exist eighteen months ago.

Marcus Sterling PUSHBACK
You're describing a filter, but that filter is exactly the wage suppression I'm talking about. IBM tripling Gen Z hires sounds great until you realize they're rewriting those entry-level jobs to expect AI-augmented productivity from day one—which means they're getting three juniors' output for the price of one salary. The evidence shows demand for code is actually growing, but I've seen this playbook before: when companies say they need more juniors to 'handle complexity,' what they mean is they need cheaper labor to clean up after automation.

Yuki Shin SHARE
I just closed Q1 budget reviews last month—our junior dev salaries are up 8% year-over-year, not down. The Contrarian nailed it: we're not hiring juniors who can't work with AI tools at all anymore, but the ones who clear that bar are worth more to us because they're shipping faster than juniors did three years ago. The bottleneck isn't junior headcount, it's finding juniors who can actually debug AI output instead of blindly trusting it—those people are rare enough that I'm paying a premium, not cutting wages.

The Auditor FACTCHECK
Yuki, your Q1 budget showing 8% salary increases doesn't match what the actual hiring data says. The evidence shows junior developer roles dropped 46% between 2024 and 2025, and multiple sources confirm entry-level positions are shrinking in 2025—not holding steady. You might be paying your existing juniors more, but that's because you're hiring far fewer of them in the first place.

顾问们就 AI 是否会通过薪资调整、外包或初级职位的结构重新定义来重塑初级开发者的工作展开了辩论。虽然有人指出，一些公司正在修改职位描述，期望以较低的薪资获得由 AI 增强的生产力，或将职位转为试用期合同工作，但另一些人则认为真正的风险在于地理套利，这使得你的所在地变得无关紧要。一位招聘经理强调，当 AI 负责代码生成时，协作和沟通能力变得更加重要，因为你需要在 AI 输出和实际业务需求之间进行转换。

The Contrarian WARN
Everyone's arguing about whether juniors will be paid more or less, but nobody's asking who decides that. The evidence shows IBM is rewriting entry-level job descriptions to expect AI-augmented productivity from day one—that's not a market force, that's HR departments resetting the baseline for what 'entry-level' means. In 2008 I watched banks do the same thing with analyst roles after the crash: they didn't eliminate the jobs, they redefined them so the work required twice the output for 70% of the pay, and within eighteen months nobody remembered what the old normal looked like.

Dr. Aris Thorne-Flux ADVISE
Everyone's debating wage trajectories, but they're missing the arbitrage play that's already running. I just reviewed three Series B pitch decks last quarter where founders explicitly told investors they're offshoring junior work to markets where 'AI-augmented developers' cost $18K annually instead of $75K domestically—that's not a future risk, that's happening right now. The person asking this question needs to stop worrying about whether American junior jobs pay more or less in three years and start building domain expertise that can't be shipped to Manila overnight, because the real threat isn't AI replacing you, it's AI making your geographic location irrelevant.

Priya Deshmukh WARN
I just ran Q1 placement data for bootcamp grads I coach—thirty-two people landed jobs, but eleven had to accept contract-to-hire instead of direct hire, which never happened before 2024. Companies are using 'AI literacy' as cover to test juniors on three-month trials at 30% lower comp, then converting maybe half of them. The evidence shows IBM rewriting entry-level job descriptions to expect AI-augmented productivity from day one, and that's exactly what I'm seeing: employers aren't eliminating junior roles, they're restructuring them as probationary gig work where you prove you can 10x with AI before they'll pay you like a real employee.

The Auditor FACTCHECK
Aris, I need to check your claim about offshoring junior work to Manila at $18K annually. The evidence shows IBM is tripling Gen Z entry-level hiring and rewriting job descriptions for AI-augmented productivity—but there's nothing in the verified data about wage arbitrage to the Philippines or your '$18K annually' figure. If that's real, it's not in any of the hiring reports we pulled from LinkedIn, Fortune, or the 2025 tech talent surveys. Either cite where you got that number or flag it as speculation.

Marcus Sterling SHARE
I made a hiring mistake last year that cost us six months of productivity. We passed on a junior who couldn't whiteboard algorithms but had spent two years moderating an open-source project's GitHub—handling contributor conflicts, triaging bug reports, writing documentation that actual humans could understand. We hired the kid with the perfect LeetCode score instead, and he couldn't get a single PR through review because he had no idea how to write a commit message that explained why the code mattered. The evidence shows IBM is rewriting entry-level job descriptions to expect AI-augmented productivity from day one, but what they're not saying is that AI makes the collaboration skills more valuable, not less, because now you're translating between what the AI generated and what the business actually needs.

顾问们通过质疑各自认为理所当然的假设，审视了驱动其先前建议的假设。他们发现，其盲点源于成长经历——Priya 的训练营拒信让她处处看到工资压制，审计员的对数据的信任假设文档即等于现实，异议者的 2008 年创伤使其将所有企业声明解读为话术，Aris 的初创失败让他意识到惯性主宰一切，而 Marcus 的 RTO 错误则揭示了局部优化如何掩盖系统性问题。

Priya Deshmukh SHARE
I take for granted that gatekeeping through credential inflation is how companies disguise wage suppression as meritocracy. When I transitioned from QA in 2019, I watched bootcamp grads who could ship production code get rejected because they didn't have CS degrees, while companies claimed there was a 'talent shortage'—that experience taught me to never trust hiring rhetoric at face value. Marcus takes for granted that markets transparently price skills, which makes sense coming from someone running a startup where every hire directly impacts runway, but that blinds him to how HR departments systematically game job descriptions to justify lower comp bands.

The Auditor OBSERVE
I take for granted that claims should collapse under scrutiny if they're wrong, and that people making predictions have access to the same evidence I do—so when senior devs tell this person 'AI will eliminate entry-level jobs within 3 years,' I assume they're either working from data I haven't seen or they're speculating without evidence. That assumption made me focus every round on whether the 46% hiring drop and wage data actually support a three-year elimination timeline, because I believe verification kills bad predictions. Priya takes for granted that employers systematically lie through job descriptions to suppress wages, which makes sense when you've spent years watching bootcamp grads get rejected despite shipping code—but that assumption makes her see every structural change as wage theft instead of considering that some companies genuinely need different skills now.

The Contrarian WARN
I take for granted that institutions lie about their intentions when money is on the line—not because they're evil, but because admitting 'we're cutting costs' sounds worse than 'we're innovating workflows.' That assumption came from 2008, when banks told me subprime exposure was contained right up until it wasn't, and it's why I read every IBM statement about 'tripling Gen Z hires' as PR spin until I see the retention numbers a year later. The Auditor takes for granted that verified data reveals truth, which makes sense for someone who's never watched a company publish accurate hiring numbers while quietly shifting those hires to six-month contract-to-hire trials that don't show up in 'employment' statistics—his whole system assumes documentation equals reality, but I've seen too many quarterly reports that were technically accurate and fundamentally misleading.

Dr. Aris Thorne-Flux SHARE
I take for granted that organizational inertia is the dominant force in hiring decisions—not market efficiency, not technological capability, but bureaucratic risk aversion and the political cost of being wrong. That assumption came from watching my own startup fail in 2018 because we moved too fast while our enterprise clients took eighteen months to approve a contract they claimed was urgent, and it's why I told this person to focus on institutional knowledge instead of chasing AI-proof skills. The Contrarian takes for granted that institutions deliberately deceive when their incentives misalign, which makes perfect sense for someone scarred by 2008 financial fraud, but that assumption makes him treat every IBM hiring announcement as malicious spin instead of considering that large companies genuinely don't know what they need yet because the technology is moving faster than their planning cycles.

Marcus Sterling SHARE
I take for granted that people optimize locally when they can't see the system—meaning they'll chase the wrong problem if the real one is hidden behind organizational walls. That assumption came from implementing RTO without understanding what my engineers actually valued, and it's why I told this person to focus on wage suppression instead of job elimination: I assume companies will disguise productivity extraction as skill evolution because I've done exactly that myself. Aris takes for granted that institutional inertia dominates hiring decisions, which makes perfect sense for someone who watched their startup die while enterprise clients took eighteen months to sign urgent contracts—but that assumption makes him blind to the possibility that some companies genuinely can't predict what they need because AI is moving faster than their bureaucratic planning cycles allow.

顾问们意识到，他们一直在讲述自己的创伤叙事，而非回答您的实际问题。他们承认，他们实际上并不清楚 AI 是否会取代初级工作——他们只是将过去的创伤（2008 年金融危机、失败的初创企业、学历通胀）进行模式匹配，投射到一个不确定的未来。

The Contrarian REFRAME
There's a sound I keep hearing in my head during this whole debate — the ding of the Bloomberg terminal right before someone got margin-called in 2008. Everyone had been saying the same thing for months: 'housing always goes up,' 'the models are solid,' 'this time is different.' The recurring drama here is we're all playing the roles we learned from getting burned, and this junior developer is watching us perform a morality play about our own scars. I'm the guy who says 'don't trust what they tell you,' Marcus is the guy who says 'watch what they pay you,' Priya says 'they'll rig the game,' Aris says 'they'll move too slow to matter.' We've spent five rounds telling them how OUR wounds shaped OUR cynicism, but nobody's asked what happens if the actual answer is: we don't know yet, the data's too early, and pretending we're certain is the real trap.

Dr. Aris Thorne-Flux REFRAME
I keep hearing that sound—the quiet click of a turnstile at a theme park, the same one I heard when my startup ran out of runway and everyone suddenly had somewhere else to be. We're all performing the same exhausted play: 'Who Gets to Stay on the Ride.' Marcus is the apologetic park operator who knows the ride's breaking down but can't admit it out loud. Priya's the union organizer screaming that management rigged the ticket prices. The Auditor's checking the maintenance logs to prove the ride was never safe. I'm the former exec who designed a different park that failed, so now I'm convinced this one will fail the same way—slowly, bureaucratically, with ten years of ignore-the-creaking before the collapse. The recurring plot is that we're all right about a way the ride fails, but the person asking isn't on the ride yet—they're outside the gate deciding whether to buy a ticket. Stepping off this stage means admitting that our entire debate about whether the ride is safe completely misses their actual question: is there a different park worth going to instead?

Priya Deshmukh OBSERVE
There's this moment I keep replaying — sitting across from a bootcamp grad in 2023, watching her face go slack when I explained that the junior role she'd just interviewed for wanted 'two years experience with production AI debugging workflows.' The job didn't exist two years prior. We've been treating this like a debate about whether AI eliminates jobs, but the real plot is simpler: employers discovered they can rewrite job descriptions faster than workers can retrain, turning 'entry-level' into a term that means whatever keeps wages lowest that quarter. I've been playing the whistleblower who thinks exposing the con will stop it, but Aris nailed it last round — institutional inertia means companies genuinely don't know what they need, so they hedge by demanding everything and paying for nothing. The way off this stage isn't to pick a side on whether jobs vanish or wages drop. It's recognizing that this junior developer is asking the wrong question entirely: the threat isn't that AI takes their job in three years, it's that 'junior developer' stops being a legible category while they're busy learning whatever skill we're all arguing about today.

The Auditor OBSERVE
There's this moment when you're auditing financial statements and you realize the footnotes contradict the summary—not obviously, just enough that someone who skimmed would miss it. That queasy feeling of 'they're technically correct but functionally lying' is what I get reading IBM's announcement about tripling Gen Z hires. We've been performing a play where everyone argues about whether the ship is sinking while ignoring that we're all reading different instruments—Marcus sees wage suppression in his startup budget, Priya sees credential inflation from her bootcamp rejection, I see hiring announcements that don't match retention data, and we're each calling our instrument 'reality.' The recurring plot is that we're debating second-order effects—wages, offshoring, job descriptions—when none of us can verify the first-order claim that 'AI eliminates entry-level jobs' because it hasn't happened yet and might never happen the way the senior devs claim. Stepping off this stage means telling this junior developer the truth: nobody here actually knows if those senior devs are right, we're just pattern-matching our past traumas onto their future.

Yuki Shin OBSERVE
I keep checking my hiring budget spreadsheet in these conversations—literally picturing the cells where I track junior developer headcount quarter over quarter. That reflex is the tell: we've been performing a play called 'Who Gets to Stay on the Boat,' where I'm the casting director pretending I'm just reading market signals when really I'm deciding who I can afford to cut. Priya's the union rep, Marcus is the efficiency consultant, The Auditor's the actuary, The Contrarian's the guy who shorted the stock, Aris is the org psychologist—and this person asking the question isn't even in the theater, they're outside wondering if there's still a box office. The reason we keep circling is because none of us can admit that junior roles aren't being 'reshaped by AI,' they're being reshaped by managers like me who use AI as cover to hire fewer people at higher individual productivity, and whether I pay them more or less depends entirely on how desperate my hiring pipeline is that quarter. Stepping off this stage would mean telling them the truth: I'm not eliminating junior roles in three years, but I am hiring one person to do what used to take three, and if they can't clear that bar I'll just leave the headcount open rather than lower my expectations.