Daily Tech News - 2026-02-14

As AI attempts to homogenize the human experience, the tech world is rediscovering that true innovation remains rooted in the non-obvious interactions of architecture and the stubborn realities of human behavior.

The industry’s current obsession with large language models has birthed a popular syllogism: since LLMs can write code, everyone will become a developer. However, this vision borders on fantasy, ignoring the fact that most humans simply aren't tinkerers. This shift toward "costumed" AI—where frozen weights and in-context learning replace true mastery—is creating a landscape of "counterfeit people," prompting urgent calls for federal laws against AI impersonation to protect human identity.

While modern tools favor abstraction, the history of computing reminds us that efficiency is often found in the "material." As Alan Kay famously noted, architecture must dominate material as complexity grows; we see this in the legacy of Michael Abrash, whose low-level optimizations once doubled the Quake framerate, and in the intricate micro-instructions of the Intel 8087 coprocessor. Even our visual interfaces are undergoing a deconstruction; while design remains wedded to the GUI, there is a growing movement toward text-driven paradigms, even as modern browsers like Safari struggle to balance new features like text-wrap: pretty with traditional typography.

Beneath these high-level shifts, the bedrock of the discipline remains mathematical and procedural. Whether it is implementing Stan Wagon’s algorithms for expressing primes as sums of squares in Python or navigating the consequential naming rules of package management, the devil remains in the details. Ultimately, technology is only as effective as its reach; whether re-launching a community three times to find an audience or battling the "oopsie" of Microsoft’s auto-renewing billing systems, the human element remains the most unpredictable variable in the machine.

Featured Articles

ai-twitter-s-favourite-lie-everyone-wants-to-be-a-developer-6058e480

Twitter's latest consensus on inevitability: now that large language models can write code, everyone will become a software developer. It's a syllogism, after a fashion, but its premise = so wildly disconnected from how actual humans behave that it borders on fantasy. Most people are not tinkerers.

Keywords: ai twitter, folks ai, developer twitter, ai write, use ai, ai use, twitter, ai, twitter building, latent creators
Source: joanwestenberg.com

as-complexity-grows-architecture-dominates-material-5686a3de

Alan Kay: As Complexity Grows, Architecture Dominates Material. Kay: Architecture is non-obvious interaction between simple materials. Chartres Cathedral contains less material than the Parthenon, despite being enormously larger.

Keywords: grows architecture, invent architecture, architecture dominates, architecture machinery, architecture, architecture built, architecture says, building thing, building, designing building
Source: worksonmymachine.ai

design-deconstruction-5011a4dc

Design is perhaps the software paradigm most wedded to the mouse and the GUI. But there’s no reason it can’t be text-driven. What if design needs to be a bit more deconstructed, where pieces are more separated out?

Keywords: design deconstruction, creativity driven, deconstruction design, creativity, designing, form creativity, graphical designed, deconstructing things, graphic design, function creativity
Source: feed.tedium.co

finding-a-non-square-mod-p-c664438c

Stan Wagon’s algorithm for expressing an odd prime p as a sum of two squares when it is possible (i.e. when p = 1 mod 4) requires first finding a non-square mod p. Wagon recommends searching for c by testing consecutive primes q as candidates. You can test whether a number q is a square mod p by computing the Legendre symbol.

Keywords: find_nonresidue legendre_symbol, computing legendre, legendre_symbol nextprime, quadratic residue, odd prime, sympy legendre_symbol, test prime, legendre_symbol, square mod, legendre_symbol returns
Source: johndcook.com

finding-a-square-root-of-1-mod-p-df51788a

The previous two posts have discussed Stan Wagon’s algorithm for expressing an odd prime p as a sum of two squares. This is possible if and only if p = 1 mod 4, the same condition on p for −1 to have a square root. If a solution x exists, how do you find it?

Keywords: odd prime, mod solution, solution mod, finding square, x² mod, squares possible, prime theorem, squares, root mod, prime
Source: johndcook.com

how-michael-abrash-doubled-quake-framerate-0b5e8e09

With the 1999 release of the Quake source code, came a readme.txt written by John Carmack. It is possible to change a #define and build with only C code, but the software rendering versions lose almost half its speed. Quake framerate is halved without Michael Abrash's optimizations.

Keywords: compilation quake, assembly quake, quake assembly, quake framerate, run winquake_no_asm, vc compiler, x86 assembly, compiler, code quake, version winquake
Source: fabiensanglard.net

instruction-decoding-in-the-intel-8087-floating-point-chip-ccde9068

The 8087 chip acts as a coprocessor with the main 8086 (or 8088) processor. When a floating-point instruction was encountered, the 8086 would let the 8087 carry out the instruction. To execute an instruction, the chip decodes the instruction and the microcode engine starts executing the appropriate micro-instructions.

Keywords: 8086 computes, complicated 8086, determine 8086, instruction 8086, 8087 instruction, registers 8086, 8086 registers, 8086 uses, 8086 instructions, 8086 processor
Source: righto.com

justifying-text-wrap-pretty-f2fb86fc

Safari is the first browser to ship a non-joke implementation of text-wrap. The combination with text-align looks ugly, as you can see in this very blog post.

Keywords: text wrap, line breaks, narrower paragraph, wrap pretty, break text, width paragraph, width greediness, line breaking, ugly typography, paragraph text
Source: matklad.github.io

launch-it-3-times-591081eb

Re-launching under a new name can be a key strategic move if you think that you’re not effectively reaching your target audience. With Glitch, that was the third name we actually launched the community under, a fact that I was a little bit embarrassed about.

Keywords: changing launching, successful launch, idea launching, launching new, launch times, company launch, launching really, launching, say relaunches, actually launched
Source: anildash.com

microsoft-game-pass-ultimate-billing-fraud-ea17664f

Microsoft Game Pass Ultimate Billing Fraud. I signed up for it and then made absolutely sure to disable any autorenewing settings everywhere I could. Somehow in the intervening three years autore new got turned back on. Oopsie whoopsie sowwy.

Keywords: billing fraud, subscription expire, cheats microsoft, trusted microsoft, fraud purchased, xbox live, purchased xbox, ultimate billing, outright fraud, fraud
Source: jayd.ml

package-management-namespaces-d1838fe9

Every package needs a name. The rules for how those names work is one of the most consequential decisions a package manager makes. I categorized the approaches previously and touched on the tradeoffs briefly.

Keywords: packages namespaces, namespaced packages, package names, namespaces rubygems, namespace package, package management, introducing namespace, discussing namespaces, package prefixes, flat namespaces
Source: nesbitt.io

the-very-old-repaint-mode-gui-approach-70c9a7f2

A modern immediate mode system generally uses double buffering. In the old days, the idea of having two buffers for your program's window was a decided luxury. You directly repainted some or all of your window any time you needed to change anything in it.

Keywords: buffering repaint, repaint mode, window repaint, selectively repaint, implement repaint, operations repaints, repainted window, minimal repainting, repainting, repainting instead
Source: utcc.utoronto.ca

tiny-corp-s-product-a-training-box-ec7d5901

If trends continue where there’s a single model with frozen weights and all learning is in-context, the cloud will win. Everyone has the same Claude/Codex/Kimi, with the same weights, the same desires, and the same biases. This isn’t learning, it's costuming.

Keywords: environments cheaper, tiny corp, models better, product training, models live, models bad, llm training, models win, models good, llms
Source: geohot.github.io

wagon-s-algorithm-in-python-45d8de95

The last three posts have been about Stan Wagon’s algorithm for finding x and y satisfyingx² + y² = p where p is an odd prime. The last post shows how to find a quadratic non-residue. This post shows a modified version of the euclidean algorithm.

Keywords: integer square, euclidean algorithm, integers square, odd prime, my_euclidean_algorithm, my_euclidean_algorithm isqrt, def my_euclidean_algorithm, return my_euclidean_algorithm, my_euclidean_algorithm stop, x² mod
Source: johndcook.com

we-urgently-need-a-federal-law-forbidding-ai-from-impersonating-humans-d0ea7e57

The late philosopher Daniel Dennett called for a law forbidding “the creation of and ‘passing along’ of counterfeit people’ The need for the kind of law he was calling for is now urgent, writes Andrew Keen.

Keywords: impersonating humans, ai impersonating, fake videos, human scammers, counterfeit people, counterfeiting, impersonating, data faked, enforce ai, pretend human
Source: garymarcus.substack.com