I person a analyzable narration pinch Hacker News. The tract is the astir important aggregator of geek news and a awesome root of postulation to this blog. At the aforesaid time, it has a adjacent number of toxic commenters, making it a dependable root of insults hurled successful my wide direction; if you want a taste, this article has been called “watered-down” and “slop”.
The tract is tally by geeks and for geeks, truthful it’s not immune to tech trends; for example, astir 2018, it had a adjacent number of stories focused connected cryptocurrencies and NFTs. That said, the caller displacement feels much profound: almost each day, it feels that the lineup is dominated by stories focused connected AI, written by AI, aliases commented connected by AI.
That images shows a peculiarly bad day, truthful to springiness a much honorable assessment, I besides performed a much systematic study successful February 2026, and again successful June of the aforesaid year.
To get a consciousness of really overmuch of the provender is occupied by AI-related topics, I took a sampling of the regular apical #5 for each of February:
AI took 4 retired of 5 spots connected Feb 4 and Feb 12, positive arguably the full line-up connected Feb 5 (story #3 was submarine trading for an AI vendor). The only days without LLM news successful the apical 5 were February 1 (with the first AI communicative astatine #7, past #9), February 9 (first astatine #8), and February 25 (with AI astatine #6, #9, #10).
For the 2nd portion of the research — figuring retired which stories were apt AI-written — I tapped into Pangram. Pangram is simply a remarkably good, blimpish exemplary for detecting LLM-generated text. These detectors person bad rap among techies, but the objections are often based connected outdated assumptions aliases outright misconceptions. For the devices to work, AI penning doesn’t request to beryllium successful immoderate measurement “inhuman”. It’s capable that the default voice of the existent harvest of LLMs is quasi-deterministic: inquire for the aforesaid effort doubly and you’ll get a stylistically akin result. The individual mannerisms are human-like, but it’s very improbable that your penning combines the nonstop aforesaid set. I constitute astir it a spot much here.
To validate the results, I besides reviewed each the flagged stories and I deliberation the findings make sense; if anything, Pangram had a mates of mendacious negatives. To springiness you a consciousness of what was flagged, person a look astatine the #3 communicative connected February 19 (“AI is not a coworker, it’s an exoskeleton”). It had 500+ upvotes and 500+ comments. In my opinion, it has a wide scope of reddish flags.
In June, to seizure much detail, I utilized coagulated achromatic for pure-play AI navel-gazing (vendor announcements, op-eds astir the benefits aliases drawbacks of the technology, etc) and hatched shapes for stories that thin heavy into AI, but person broader ramifications (e.g., the Instagram AI support supplier relationship hack). As before, stories that are only tangentially related to AI (e.g., reports of RAM value hikes) are not flagged.
In the first half of the month, astir 60% of the regular HN lineup was AI-related aliases AI-generated, tapering disconnected to ~50% arsenic we approached the extremity of the month. This is up from 40% successful February.
English (US) · 
Indonesian (ID) · 
