Lightnews — Scholar-powered news

Reposted by John Allsopp

John McBride

@johncodes.com

Wrote about the sheer amount of software that is going to be created: the software Cambrian explosion is upon us. Prepare accordingly.

johncodes.com/archive/2026...

The software Cambrian explosion.

johncodes.com

January 11, 2026 at 6:35 PM

Reposted by John Allsopp

Timchat Torah

@landbeforetim.bsky.social

The Cardiff Lions do an annual Drag Rugby charity match for children’s charities and it’s as beautiful as you imagined

January 10, 2026 at 7:45 PM

Reposted by John Allsopp

qustrolabe

@qustrolabe.bsky.social

haha linus vibecoding

January 10, 2026 at 10:00 PM

Reposted by John Allsopp

Paul Irish

@paul.irish

Been loving NVIDIA's NeMo models recently to complement Whisper for transcription, but the startup felt a bit heavy.

So I went on a profiling adventure!
github.com/NVIDIA-NeMo/...

Turned out to be fruitful… stoked to land a big optimization in NeMO.
github.com/NVIDIA-NeMo/...

A flame chart of loading NVIDIA's flagship ASR model, Canary. Captured with viztracer and shown in trace.cafe. It shows the load takes 52 seconds only for 3s of inference! Several annotations point out some opportunities for optimization (redundant tar extract, config read/write thrashing etc)

January 10, 2026 at 10:10 PM

Reposted by John Allsopp

Jono Alderson

@jono.id

The middle of the market is quietly dying.

Breadth and “good enough” used to work because humans couldn’t see everything.

Agents can.

If you’re not the safest default or clearly the best, you’re just noise.

www.jonoalderson.com/conjecture/t...

The middle is a graveyard

Why breadth, mere adequacy, and the “middle of the market” fail in an age of omniscient, agent-led decision-making.

www.jonoalderson.com

January 10, 2026 at 9:02 AM

Reposted by John Allsopp

Data & Society

@datasociety.bsky.social

Anthropomorphizing language conceals the limitations of AI, promoting misplaced trust. @emilymbender.bsky.social & @nannainie.bsky.social suggest focusing on a system’s functionalities: instead of saying a model is “good at” something, say what it is “good for." www.techpolicy.press/we-need-to-t...

We Need to Talk About How We Talk About 'AI' | TechPolicy.Press

We share a responsibility to create and use empowering metaphors rather than misleading language, write Emily M. Bender and Nanna Inie.

www.techpolicy.press

January 8, 2026 at 8:14 PM

John Allsopp

@johnfallsopp.com

I don’t recall talk of a ‘smartphone bubble’ (this. It by way of saying ‘therefore talk of ai bubble is wrong’ and reasoning by analogy is fraught. But I think it could be interesting to consider why.

January 8, 2026 at 9:00 PM

Reposted by John Allsopp

Steve Klabnik

@steveklabnik.com

Getting started with Claude for software development: steveklabnik.com/writing/gett...

Getting started with Claude for software development

Blog post: Getting started with Claude for software development by Steve Klabnik

steveklabnik.com

January 7, 2026 at 6:20 PM

John Allsopp

@johnfallsopp.com

In my ‘stack collapse’ piece late last year I suggested the parts of the front end stack that exist for developer convenience will come under pressure. This is Tailwind becoming commoditized by LLMs writing it. Next they skip it entirely

webdirections.org/blog/stack-c...

github.com/tailwindlabs...

feat: add llms.txt endpoint for LLM-optimized documentation by quantizor · Pull Request #2388 · tailwindlabs/tailwindcss.com

Add /llms.txt endpoint that serves a concatenated, text-only version of all Tailwind CSS documentation pages optimized for Large Language Model consumption. Extract text from MDX files, removing J...

github.com

January 8, 2026 at 12:39 AM

Reposted by John Allsopp

conputer dipshit

@davidcrespo.bsky.social

wild. Tailwind Labs revenue is down 80% due to LLMs making docs site visits unnecessary (and, I would add, making the HTML templates they were selling a lot less useful)
github.com/tailwindlabs...

@adamwathan
Member
adamwathan
commented
18 hours ago

I totally see the value in the feature and I would like to find a way to add it.

But the reality is that 75% of the people on our engineering team lost their jobs here yesterday because of the brutal impact AI has had on our business. And every second I spend trying to do fun free things for the community like this is a second I'm not spending trying to turn the business around and make sure the people who are still here are getting their paychecks every month.

Traffic to our docs is down about 40% from early 2023 despite Tailwind being more popular than ever. The docs are the only way people find out about our commercial products, and without customers we can't afford to maintain the framework. I really want to figure out a way to offer LLM-optimized docs that don't make that situation even worse (again we literally had to lay off 75% of the team yesterday), but I can't prioritize it right now unfortunately, and I'm nervous to offer them without solving that problem first.

@PaulRBerg I don't see the AGENTS.md stuff we offer as part of the sponsorship program as anything similar to this at all — that's just a short markdown file with a bunch of my own personal opinions and what I consider best practices to nudge LLMs into writing their Tailwind stuff in a specific way. It's not the docs at all, and I resent the accusation that I am not disclosing my "true intentions" here or something.

This feature is so that people can build MORE things with Tailwind in a FASTER and more EFFICIENT capacity.

@mtsears4 Tailwind is growing faster than it ever has and is bigger than it ever has been, and our revenue is down close to 80%. Right now there's just no correlation between making Tailwind easier to use and making development of the framework more sustainable. I need to fix that before making Tailwind easier to use benefits anyone, because if I can't fix that this project is going to become unmaintained abandonwar…

January 7, 2026 at 10:23 PM

Reposted by John Allsopp

Ed

@ed3d.net

I've been consciously trying to do more retweeting of folks with Good And Thoughtful Takes around AI and the like lately, so here's a starter-pack thing I'll definitely not forget to keep updated later 🫡

January 3, 2026 at 3:30 AM

Reposted by John Allsopp

Simon Willison

@simonwillison.net

Here's my enormous round-up of everything we learned about LLMs in 2025 - the third in my annual series of reviews of the past twelve months
simonwillison.net/2025/Dec/31/...
This year it's divided into 26 sections! This is the table of contents:

The year of “reasoning”
The year of agents
The year of coding agents and Claude Code
The year of LLMs on the command-line
The year of YOLO and the Normalization of Deviance
The year of $200/month subscriptions
The year of top-ranked Chinese open weight models
The year of long tasks
The year of prompt-driven image editing
The year models won gold in academic competitions
The year that Llama lost its way
The year that OpenAI lost their lead
The year of Gemini
The year of pelicans riding bicycles
The year I built 110 tools
The year of the snitch!
The year of vibe coding
The (only?) year of MCP
The year of alarmingly AI-enabled browsers
The year of the lethal trifecta
The year of programming on my phone
The year of conformance suites
The year local models got good, but cloud models got even better
The year of slop
The year that data centers got extremely unpopular
My own words of the year
That’s a wrap for 2025

December 31, 2025 at 11:54 PM

Reposted by John Allsopp

Ben Schwarz

@benschwarz.bsky.social

People like to say “once the single page app is booted, it’s faster”, but they’re wrong — stats show few subsequent navigations occur.

React simply isn’t worth the cost for most websites.

Stoyan Stefanov @stoyan.me · 11d

Web Performance Calendar day 31 article 5/5: @infrequently.org investigates SPA usage patterns showing that users often trigger only ~1 soft navigation per page load. If the heavy upfront JS costs aren’t being amortized, what are SPA's performance benefits?
calendar.perfplanet.com/2025/the-cur...

The Curious Case of the Shallow Session SPAs

Buried at the end of this year's instalment of my semi-annual series on network and device reality is a mystery: multiple, independent data sets from the Web Performance community indicate sites built...

calendar.perfplanet.com

December 31, 2025 at 10:21 PM

Reposted by John Allsopp

Alex Russell

@infrequently.org

New year, new web performance mystery (and it's a huge one):

calendar.perfplanet.com/2025/the-cur...

December 31, 2025 at 8:29 PM

Reposted by John Allsopp

Citizen Platano 🇵🇷

@daniloc.xyz

so Nemotron Nano 3 is a shockingly good model for local inference

getting roughly 90tok/s on M4 Max, with output that’s at least superficially as sharp as anything you see from Claude, seemingly quite flexible for different kinds of tasks

nvidianews.nvidia.com/news/nvidia-...

NVIDIA Debuts Nemotron 3 Family of Open Models

The Nemotron 3 family of open models — in Nano, Super and Ultra sizes — introduces the most efficient family of open models with leading accuracy for building agentic AI applications.

nvidianews.nvidia.com

December 27, 2025 at 9:20 PM

Reposted by John Allsopp

The Australia Institute

@australiainstitute.org.au

"In Australia, we subsidise the fossil fuel industry. We give more than half of the gas we export away for free, and we charge our kids a fortune to go to uni."

See more of Richard Denniss on Q&A here: https://www.youtube.com/watch?v=CAYxzRNfC6k

December 27, 2025 at 10:25 PM

John Allsopp

@johnfallsopp.com

For the 30th anniversary I was privileged to be involved in this
Project at CERN to help resurrect the original browser

www.zdnet.com/article/cern...

December 25, 2025 at 11:40 PM

Reposted by John Allsopp

depths of wikipedia

@depthsofwikipedia.bsky.social

it's Christmas Eve but it's also the 16 year anniversary of the time an art historian for the Hungarian National Gallery spotted a long lost painting while watching Stuart Little with his 3-year-old daughter, recognizing it in the movie set

December 24, 2025 at 11:40 AM

Reposted by John Allsopp

Armin Ronacher

@mitsuhiko.at

A personal recap of 2025. It was a year of a new style of engineering. Left some ideas for what might come next. lucumr.pocoo.org/2025/12/22/a...

A Year Of Vibes

A personal recap of 2025 and a year of a new style of engineering.

lucumr.pocoo.org

December 22, 2025 at 10:08 AM

Reposted by John Allsopp

AJ Fisher

@ajfisher.social

Some other teaser predictions:

- HR will be sitting in the hotseat during board meetings
- Disposable Software arrives that business users will love and IT will hate.
- The China manufacturing shock continues but this time with autonomous robots.

#hr #ai #work #strategy #business #china #robotics

December 22, 2025 at 2:00 AM

Reposted by John Allsopp

AJ Fisher

@ajfisher.social

As we come into 2026, my hot take is that Replit, Cursor etc get acquired, crushed or sold for parts.

Bundling will make a comeback as MS just make "Agentic Edit" part of a standard M365 or Github license.

This and more ajfisher.me/2025/12/22/a...

#ai #predictions #blog #business