Why LLMs aren't a dumb fad like crypto, but the biggest thing since the World Wide Web, and on the coding front, the biggest thing since IDEs and Stack Overflow
beyond the shift in market sentiment. comparing the two makes no sense when you think about goals + end consumer https://twitter.com/... @hunterwalk : after listening to people pump crypto and the VR metaverse, i'm so happy the next trend (AI) is a truly interesting and valuable technology. https://about.sourcegraph.com/ ... Ben Parr / @benparr : My favorite quote from @steve_yegge's excellent piece on how AI is transforming engineering. A must read: https://about.sourcegraph.com/ ... https://twitter.com/...
Sourcegraph BlogSteve Yegge
Context & Ripple Effects
Published days after a companion piece arguing LLMs would demolish barriers to software creation and unleash the next technology cycle, this Sourcegraph essay lands at the moment market sentiment pivoted from crypto and metaverse fatigue toward AI — and it stakes a precise claim: for software work specifically, LLMs rank alongside IDEs and Stack Overflow as an inflection point. Weeks earlier, the profile of Emily M. Bender had supplied the academic counterpoint on what these systems can and cannot do, so the essay is also a shot in that running debate.
The coding half of the thesis has since become empirically checkable: Stack Overflow's own surveys now show 84% of developers using or planning to use AI tools while trust in AI accuracy fell to 33%, and the platform's December 2025 question volume sank to 2008-era levels even as annual revenue doubled to $115M on enterprise tools and AI licensing.
First-order effects
Developers choosing whether to adopt AI assistants get a framing argument from a developer-tooling vendor — Sourcegraph — that positions LLMs as durable infrastructure rather than a speculative cycle like crypto.
The IDE-and-Stack-Overflow comparison puts Stack Overflow itself directly in scope: the platform cited as the prior coding inflection point is watching AI tools absorb the question-answering role it owned.
Second-order effects
Stack Overflow's response, visible in its own numbers, is to monetize the shift: revenue doubled to $115M through enterprise tooling and AI licensing of its content even as monthly question volume fell back to 2008 levels.
If the adoption-trust gap persists — near-universal usage alongside collapsing confidence in accuracy — developer workflows institutionalize around verification-heavy AI assistance as a steady state rather than a transitional one.
Karpathy's observation that LLMs disproportionately benefit individuals more than corporations or governments points to bottom-up diffusion through practicing engineers first, reshaping team structures before org charts catch up.
The trend: Industry debate is reclassifying LLMs from a speculative asset comparable to crypto into a foundational engineering layer, with coding productivity serving as the earliest proving ground.
a delightfully terse summary of how we got to the LLMs we're all so excited about today by @Steve_Yegge https://about.sourcegraph.com/ ... https://twitter.com/...
I've just posted my first update since joining Sourcegraph six months ago, “Cheating is All You Need”. In it I introduce AI-backed coding assistants, which are the biggest change I've seen in 35 years as an engineer. Enjoy! https://about.sourcegraph.com/ ...
yegge's piece, as my colleague articulates with some colorful language, is something everyone should read https://about.sourcegraph.com/ ... https://twitter.com/...
your #1 concern when squinting at the AI future is a familiar one, who is deriving value from your data and why is it not you https://about.sourcegraph.com/ ... https://twitter.com/...
in which Steve Yegge does tech marketing. He just joined a code search company and the way to add value to LLMs that write code is with code search. Build a moat. Ride the volcano. Read the Yegge piece. @sourcegraph is in the game. https://about.sourcegraph.com/ ...
i keep thinking about how ai's hype is very different from crypto's hype — beyond the shift in market sentiment. comparing the two makes no sense when you think about goals + end consumer https://twitter.com/...
My favorite quote from @steve_yegge's excellent piece on how AI is transforming engineering. A must read: https://about.sourcegraph.com/ ... https://twitter.com/...
Beautifully put. I'm only going to use the LLM that best understands my codebase and what I want to do in it! “If you can't feed the LLM your whole code base, and you can only show it 100k characters at a time, then you'd better be really goddamn good at https://about.sourcegraph…
after listening to people pump crypto and the VR metaverse, i'm so happy the next trend (AI) is a truly interesting and valuable technology. https://about.sourcegraph.com/ ...
“LLMs aren't just the biggest change since social, mobile, or cloud-they're the biggest thing since the World Wide Web. And on the coding front, they're the biggest thing since IDEs and Stack Overflow, and may well eclipse them both.” - Steve Yegge https://about.sourcegraph.com/ …
I feel like this post plus Wolfram's super long one should get you to a spot where you see the potential. It won't be in interacting with the LLM directly - it'll be tailoring both ends to a specific outcome. (Which is obvious, but only if you dig and pay attention) https://twitt…