Vitaly Friedman's conversation with Tejas Kumar for the Smashing Podcast opens on the question of whether technology is making developers redundant, and whether AI will soon replace them. Kumar, who started writing code at eight and has since held roles at G2I, Vercel, Spotify and Xata, frames his answer around the idea that machines are currently being designed as helpers rather than replacements.
Before the AI question, the episode traces how he arrived at the craft. A childhood illness limited what he could physically do — he could not attend school daily, carry a backpack, open doors or walk upstairs — so typing on a keyboard was one of the few things available to him. His family was poor but owned a computer. Coding, in his words, found him. The draw was JavaScript's immediate feedback loop: a line typed into a browser console and executed. Photoshop came first, including tutorials for the shiny Aqua-era UI balls and Tux penguins, then the slicing tool that produced HTML, then JavaScript once the question became how to make the result interactive. The appeal was creating things with no physical cost and no financial barrier beyond the hardware.
Friedman names the likes of Lex Fridman to make the point that holding strong opinions is fine as long as you are willing to be argued out of them. Kumar agrees that the underlying skill is critical thinking, and describes his own rule as speaking the truth in love or kindly — strong opinions held under an umbrella of kindness, because nobody listens to an angry person.
Kindness and compassion is the most important.
Maturity, and what it costs
Kumar pushes back gently on Friedman's claim that the industry is mature. He associates maturity with the fear he had as a rebellious young person — that adulthood meant a suit and the loss of childlike wonder — and he sees the same trade in today's web. GeoCities and Angelfire pages, under-construction banners, the dancing baby, a clock cursor that followed you around: gone, he says, not because they are impossible but because they are no longer cool. His own site is an attempt to bring that back. Drag the cursor and it cycles through random photos of him — a flower costume, a shirtless muscular body with his face on it — added by the community through pull requests, since the site is open source. Friedman connects this to Smashing's cats: non-essential, and essential precisely for the whimsy.
Autonomy, trust and boundaries
Asked what he took from each employer, Kumar points to urgency, ownership and autonomy as the combined lesson of Xata and G2I. He recalls an old Steve Jobs interview on people leaving when they cannot have ideas, and describes reporting to a CTO at G2I who would align on a plan on Monday, say "I trust you" and "make it happen", and expect to be contacted only if needed. Later, as director of developer relations at Xata, he carried that pattern into his own team: plan on Monday, check in on Friday, come to me in between if you need anything. The result, he says, was urgency without stress.
Friedman brings up the "manual for me" practice, in which every employee documents how they work best — calendars, preferences, whether morning interruptions are acceptable — and links it from a Slack profile, something he sees as particularly valuable in large remote organisations. Kumar says G2I ran the same thing under the name social contracts. On remote versus office, he is not rigid: office work would be fine if other factors lined up, and he cites a friend's preference for having the choice — remote for a season, onsite for a season.
Recessions, specialisation and who gets left out
On the current wave of layoffs, Kumar is unsentimental: GDP curves dip, recessions are normal and predictable, and layoffs follow. They are not good — they affect families, and he feels the squeeze of unemployment himself after having quit — but they are expected. His expectation for AI is that it helps rather than replaces us, with the caveat that giving ChatGPT internet access it does not currently have would add superpowers and make it more threatening; he also notes that there are humans with a vested interest in keeping humans employed. What he finds genuinely good about the industry's present state is its branching structure. In his early days there was no split between front end and back end, only the webmaster who designed, developed and dragged files over FTP to a shared host. That gave way to front end, back end and DevOps, then machine learning engineers and data scientists, and DevRel, which he considers still young. More specialisations mean more ways into the field. Since everything is turning into tech — clothing stores into e-commerce, travel agents into ticketing sites — and since the original arrangement of the world was heavily unequal, he argues that campaigning for fairness, diversity, equality and inclusion matters to stop people being left out.
Friedman offers the counter-intuition that speed has never actually reduced the work: technology promised less, and instead people do more, with fragmented attention rather than genuine multitasking. He does not believe technology is making us redundant, only changing what we do; Kumar agrees.
Where the frameworks stand, and why serverless unlocked them
Pressed on whether React will still matter in twenty years, Kumar guesses he will be proven wrong in one. His starting observation is that the web is held up by underappreciated giants: jQuery and WordPress power more of the internet than React does. There was a period when React had enough momentum to look cooler than jQuery, and jQuery's usage held even as its perceived value fell. He sees 2023 as the start of a similar turn for React — heavily used, but with diminishing perceived value, especially against Solid and Qwik, which avoid shipping a whole virtual DOM implementation to the browser and so are objectively faster and lighter. For some that difference is mere milliseconds; he cites Google and Chrome work showing that milliseconds make millions. React is unbeatable, he argues, in the same sense that jQuery and WordPress have not been beaten — not least because teams at Google invest in React and in how it runs in Chrome. His broader claim is that predicting a framework's direction means looking at the adjacent infrastructure: Vercel, Netlify and Cloudflare made deploying a Node server easy and produced what he calls server liberation. Before them, server rendering was awkward because servers were hard to configure for client-side apps — you needed a 404 rewrite to hand over your index.html so the client could download. Once servers became easy, server rendering became viable, and React, Solid and the rest shifted accordingly.
Conferences, online and in person
Kumar confirms he is working on a conference, framing it as joining the three perspectives of attendee, speaker and organiser, and as connecting sponsors to events that lack funding, so attendees get the best possible experience. He is certain in-person events are back for good, and equally certain that nothing replaces face-to-face interaction — no screen substitutes a warm, flesh-and-blood person in front of you. Online does not disappear either: organisations already invested in streaming, it sells tickets by volume, and it includes people who cannot travel, so it persists as a fallback track alongside the room. On mistakes he has seen organisers make, micromanagement is the pattern he would most want to avoid repeating from past employers, and the conference error he singles out is overselling the venue and failing to tell speakers they will not be recorded — by which point the effort of travelling is already spent.
What comes next
At the time of recording Kumar was between jobs, having decided to spend the week working out his next move. He expects to keep making YouTube content, and says he is leaning towards starting some kind of enterprise around DevRel — a place where people can be creative and do their best work while earning well. He labels the fallback plainly: a job at a tech company on a decent salary, in a field he cares about. Large tech empires do not attract him; Spotify, at around 4,000 employees, was the closest he has come to that energy and the experience told him it was not for him. He also describes meeting load: the pandemic pushed Spotify, previously never remote, into remote work with async not yet established, so days filled with video calls. Friends at Google and Meta show him calendars he does not envy. His lesson from all of it is that people — himself included — are poor at setting workplace boundaries, and that employers who suggest no-meeting days and establish boundaries on their people's behalf would help. He is currently at work on three projects of his own: one secret, described only as very social; the DevRel consultancy; and his YouTube channel, which he has wanted to do for years and wants to do well rather than averagely. On what he is learning lately, it is communication: a high-energy "what's up everybody?" lands on a conference stage at Smashing Conf New York or Freiburg and reads as strange on YouTube, because different modalities have different preferences. He has yet to crack TikTok, joking that it would be mostly dance.
His closing advice is about kindness and compassion.
