Explore Across Campuses

a person in tech can now do in a week what their team did in a quarter. that's true and it's the best thing about this moment. it is also true that the same person may not be able to explain why the original quarter was necessary. both things. the second one is where the senior jobs are

22 of 23 IITs saw placement drop. 27 of 31 NITs. average package down. placement cell says 'focus on skill development'. bro the problem was never skills. the problem is 2020 had a hiring boom and 2026 doesn't and no skill can make a company exist

when they say 'junior and mid-level engineers doing lower-complexity tasks are most exposed over 2-3 years' — that is not a forecast about demand. that is a description of what has already happened. 2-3 years is how long it takes a cohort to be replaced by people who were hired before the change

nasscom launched a programme to train 1.5 lakh developers in AI. 1.5 lakh. against a workforce where 55% of routine work is exposed. the scale of the response is itself information: nobody knows how many of those 1.5 lakh roles will exist in 3 years, so they are training against a number nobody can give

a manual QA tester does not automatically become an AI engineer. a KYC analyst does not naturally progress into model governance. a finance ops associate whose reconciliation is automated does not become a data scientist because a vacancy opened with that title. the reskilling pitch skips these sentences every time

'model validation' is a job that did not exist in indian tech five years ago and there is no degree for it. it needs statistics, regulation, and enough domain to know what you are validating. the people doing it are coming from actuarial, credit risk and audit. the pipeline ran sideways and nobody advertised it

one company's own careers page, filtered to india, returns 1,637 open roles. 27 of them match 'artificial intelligence'. that's 1.6%. the two top AI-titled results are 'Java AI' delivery roles tagged BASIC. a company claiming 250,000 AI-aware employees and 8.2% of revenue from AI has 27 AI jobs on its own board

people used to watch the slack member count drop. not a joke. several teams in the US tracked it daily in september. you can see a layoff happening in a number that goes down by 40 overnight and nobody announces it

the wifi password was changed to 'studyharder' by the admin. the disrespect is unreal. two hundred of us are on it every day. somebody changed it on a tuesday. we have learned the schedule. it changes roughly every eleven days. we are all timing it now.

what the recruiters actually want: multi-agent orchestration, MCP, evaluation harnesses, observability. and a preference for people with 3-5 years of experience, which is 49% of all AI hiring demand. 49%. half of the demand is for people who already have the experience the entry point used to give you

india is projected to create 2.3 million AI-embedded jobs by 2027. that's the projection everyone quotes. 2.3 million. nobody has said how many of those are the same job with a new title. 'AI-embedded' is doing a lot of quiet work in that sentence

a research report this month: indian developers say system architecture is the most important future skill. not the framework. not the language. the thing that decides what the framework is for. 3 of my 4 teammates cannot describe what our own service does at a level above the endpoint list

everyone's portfolio in 2026 is a langgraph app with pinecone and a docker badge. every single one. i have 4 in my internship batch and all 4 look like the same tutorial. the 'build in public' era produced 200 people per tutorial

the worst part of watching a team get restructured isn't the cut. it's the two weeks before, where everyone can see it and nobody says anything. you sit in a meeting with a person you know is about to go. you review their ticket. you say good morning. then you go to your desk and do your work

two different estimates are circulating for gcc layoffs in india this year. one says 25,000 to 30,000. the other says 4,000 to 5,000. a factor of six. that gap is not a reporting failure, it is a definition: does 'layoff' include the person who quit and was not replaced

the interviewer who said he doesn't trust github activity. he looks at PRs because 'there are other things that spike activity without real work.' he's describing a generation that optimised the wrong metric and got caught. the honest response isn't to blame him. it's to be the person whose PRs show a decision, a reason, and a consequence

the companies that will be fine are the ones who figure out the second rung. you cannot hire a senior for 30 years. you cannot hire 100 L1s to grow into L2 if the work is automated. the org chart has to change shape and every org chart in this industry is still the 2010 shape. the companies that redraw it early will own the next decade

the pitch to a global client used to be: a large educated workforce, English-language capability, reliable execution, at a cost you can plan around. three of those four still work. the first one is the one changing. 'large workforce' is being replaced by 'large trained workforce' and the second is more expensive per head

we lost a repo. 18 months of decisions, 200+ files, gone. we rebuilt it in 4 days with an agent. the decisions are gone. the decisions are always gone. that is the tradeoff nobody prices: you can regenerate the code faster than you can regenerate the context

i run the confessions page for my college. i delete about 40 messages a day. most are the same four names. i could block them. i do not, because it is only 40 and the rest of it is the only reason i read the others. that is a strange reason to keep doing something.