Kore D
(He/Him)
India
Covers:
Artificial Intelligence, Cybersecurity, Cryptocurrency, Startups, Technology, Hardware, Software, Gaming
$ decoding what actually matters in tech
// AI, chips, software, security, startups
original analysis, no slop
read the full breakdown ↓
Interview
What was your first job as a journalist?
This one. I came to journalism from web development, not a newsroom. GENZ TECH is my first publication, launched in 2026, and the developer background is why it works: when a company publishes a benchmark, I don't just quote it, I can check it.
How is social media changing news?
The story now breaks in a screenshot and gets corrected in a comment thread. For my beat the interesting shift is that verification moved to the replies: a lab posts a benchmark on X, and within hours someone has re-run it and posted the real number underneath. Half my job is reading the replies before believing the post.
What does it mean to be a journalist?
I'm new enough to this that I hold the definition loosely. But the working version I use every day: a journalist is someone who checks. Anyone can repeat an announcement. The job starts when you ask who measured it, how, and what happens when someone else runs the same test.
What's the funniest news-related #hashtag you've seen?
#FundsAreSafu. Born from an exchange hack, immortal ever since. Nothing says confidence like a misspelled reassurance that appears only when funds are, in fact, not safu.
What tools and software do you use to do your job?
The site is custom built, which is half the fun of running it. Writing happens in [your editor], the interactive tools (AI coding leaderboard, cost calculator, CVE watchlist) are my own code, and the data behind them comes from independent eval sources like vals.ai plus primary filings and reports. Beyond that: RSS for monitoring, [X/Twitter lists, Google Alerts, whatever you use] for catching stories early, and a long spreadsheet of claims to re-check.
Who do you wish followed you?
The people who run independent evals: vals.ai, security researchers, whoever re-tests vendor numbers. They're my beat's real editors, and I'd rather have fifty of them than fifty thousand casual followers. Also anyone with an unpublished benchmark discrepancy and a spreadsheet.
What story are you most proud of writing or working on?
My coverage of Kimi K3. The easy story was "third best coding model at a third of the price." The better story was that Moonshot kept calling it the largest open model ever while shipping no downloadable weights, and that its own Terminal-Bench claim came in 3.3 points high when someone independent ran it. Holding both of those in the same article, that it's genuinely impressive and that the marketing doesn't survive contact with verification, is the kind of work I want the site to be known for.
What's the best pitch you ever got?
A researcher who sent three sentences: here's what we found, here's the dataset, here's why the obvious interpretation is wrong. No adjectives, no "synergy," just a claim and the evidence. That became a story the same week.
What's the worst pitch you ever got?
The site is young, so my inbox hasn't seen true horrors yet. But the fastest way to end up in my trash is a benchmark chart with no methodology. If your model is "best in class," tell me who ran the eval, on what, and where I can reproduce it. I've seen vendor benchmark claims fall apart under independent re-runs, and that becomes the story.
When you're not at a computer, where are you most likely to be?
Realistically, reading the same news on my phone. On a good day, [your actual thing: on a cricket ground, in the kitchen, on a motorcycle, hiking]. I try to keep at least one hobby that cannot ship a firmware update.
Aside from your own, what's your favorite publication to read?
SemiAnalysis for the chip supply chain, and Simon Willison's blog for AI tooling. Both do the thing I aspire to: primary sources, receipts, and no press-release paraphrasing.
What's the most common misperception about your beat?
That benchmark numbers are measurements. Most of the scores you see in launch posts are marketing that happens to have a decimal point. The number that matters is the one an independent third party gets when they run the same test, and it's routinely lower than the slide said.
Would you like to take the Muck Rack Interview?
Create a Muck Rack profile.