apike - Network

The Periodic Table of Elements, according to GPT-4o's new, smarter image generation

The Periodic Table of Elements, according to GPT-4o's new, smarter image generation

Maybe for a while the approach will be to adopt the newest minor version of whatever major version your LLM knows about. Over time “Can...

Maybe for a while the approach will be to adopt the newest minor version of whatever major version your LLM knows about. Over time “Can use new docs to work with dependencies that have had major changes since training” should become a basic competency for model training

When you pull in a dependency, there's a little trepidation when it's just had a major release. Do you start with well-tested...

When you pull in a dependency, there's a little trepidation when it's just had a major release. Do you start with well-tested NeatLibrary 2.6.11, or skip to NeatLibrary v3.0.1, with its early bugs and docs, but never need to migrate?

While I've always leaned towards just starting on the new thing, Cursor makes this tradeoff less appealing:
1. It will understand the before-training-cutoff version's features and syntax better
2. It makes migrating to new dependencies later on way easier

JavaScript Fatigue Strikes Back: Why we still have so many JS frameworks, and why maybe – maybe – we have enough...

JavaScript Fatigue Strikes Back: Why we still have so many JS frameworks, and why maybe – maybe – we have enough now?

https://allenpike.com/2025/javascript-fatigue-ssr

Initial impressions of GPT-4.5 is that it's noticeably more “personable” – easier to prompt into behaving more idiosyncratically...

Initial impressions of GPT-4.5 is that it's noticeably more “personable” – easier to prompt into behaving more idiosyncratically and less like a wikipedia page.

I've put a little 4.5-powered "product research agent" in the corner of https://forestwalk.ai/ if you're curious how this can feel in practice.

Remaining RVSPs are now open for next Tue 4's Infer Vancouver event, featuring lessons learned scaling LLMs from the team at Ada, and...

Remaining RVSPs are now open for next Tue 4's Infer Vancouver event, featuring lessons learned scaling LLMs from the team at Ada, and detecting social engineering from Ian MacKinnon: https://lu.ma/brrgkhgv

Local chain Save-On-Foods has new in-store flyers promoting their Canadian ownership. Where else can you fill your cart with pumpkin-sized...

Local chain Save-On-Foods has new in-store flyers promoting their Canadian ownership. Where else can you fill your cart with pumpkin-sized apples, giant jars of demented pasta sauce, lettuce that exceeds the boundaries of your cart, and a bunch of mysterious crimson… mush? And don't forget their patented “BEEFY DAIRY”

Our next Infer Vancouver event is in 3 weeks, featuring Ada's @davidhariri and Stingray's Ian MacKinnon – two founders working on...

Our next Infer Vancouver event is in 3 weeks, featuring Ada's @davidhariri and Stingray's Ian MacKinnon – two founders working on the forefront of building with LLMs. https://lu.ma/brrgkhgv

@atpfm You’ve discussed before how HDMI-CEC is famously fragile and hard to troubleshoot. This divides the world into “CEC unicorns”...

@atpfm You’ve discussed before how HDMI-CEC is famously fragile and hard to troubleshoot. This divides the world into “CEC unicorns” who enjoy a temporary device and input nirvana, folks who can tolerate jankiness, and those who toss it all in the sea.

It seems like a “CEC 2” that was more rigorous and configurable could be a selling point for future devices. Does John have a sense why a better CEC spec has never come about? #askatp

Sam Julien’s Infer talk is now also live, where he shares lessons learned at Writer on how to get modern AI applications shipped in the...

Sam Julien’s Infer talk is now also live, where he shares lessons learned at Writer on how to get modern AI applications shipped in the enterprise. Thanks to all our speakers, volunteers, and attendees – looking forward to the next one! ⭐️

https://www.youtube.com/watch?v=Gnoo3LH2d1o

Swyx’s talk from the NeurIPS Infer Vancouver is now live! He shares his view on the state of AI engineering, gives a demo of realtime AI...

Swyx’s talk from the NeurIPS Infer Vancouver is now live! He shares his view on the state of AI engineering, gives a demo of realtime AI APIs, and answers folks’ burning questions. https://www.youtube.com/watch?v=Hx_0TBOV0do

If you work on the web you probably have picked up that animating the height of an element is slow and worth avoiding, but this is a fun...

If you work on the web you probably have picked up that animating the height of an element is slow and worth avoiding, but this is a fun little dive into why: https://www.granola.ai/blog/dont-animate-height

Meta has a few reasons for publishing Llama with open weights, which has been great for developers and startups. But a subtle one is its...

Meta has a few reasons for publishing Llama with open weights, which has been great for developers and startups. But a subtle one is its effect on future acquisitions.

If a startup depends on Llama-based LLMs – which are by license not usable by Apple, Microsoft, Google, or Amazon – then Meta is the only Big Tech company that can acquire their technology. Bonus for Meta.

Luckily model lock-in is pretty low so far, but it's an interesting potential dynamic.

The Era of Tab Continuation: A delightful new UX pattern for productivity apps.https://allenpike.com/2025/tab-continuation-completion

The Era of Tab Continuation: A delightful new UX pattern for productivity apps.

https://allenpike.com/2025/tab-continuation-completion

ChatGPT advanced voice mode is helpful for talking you through stretches! It can suggest different positions, explain the muscles involved,...

ChatGPT advanced voice mode is helpful for talking you through stretches! It can suggest different positions, explain the muscles involved, and give tips on technique.

Just be aware that if you have it describe lying down or getting on your knees, the response might be deemed too spicy

On a day to day basis, founders typically focus on what's immediately ahead – the daily process of moving the business forward.This...

On a day to day basis, founders typically focus on what's immediately ahead – the daily process of moving the business forward.

This can make it easy to lose sight of what you've achieved. For example, according to LinkedIn, we have achieved a YoY growth rate exceeding two billion percent

It's a new year, and you know what that means: time to confirm your website is still readable in Netscape® Navigator 4

It's a new year, and you know what that means: time to confirm your website is still readable in Netscape® Navigator 4

For years, I resisted creating an email list. “Does anyone actually want to read blog posts in a mail client?”But as the list has grown,...

For years, I resisted creating an email list. “Does anyone actually want to read blog posts in a mail client?”

But as the list has grown, I've noticed something delightful: I get 10-100x the response rate for folks who receive articles by email! Useful feedback, typo corrections, notes of appreciation. Well worth the effort of setting it up.

https://allenpike.com/follow/

An Unreasonable Amount of Time: A method for doing the impossible.https://allenpike.com/2024/an-unreasonable-amount-of-time

An Unreasonable Amount of Time: A method for doing the impossible.

https://allenpike.com/2024/an-unreasonable-amount-of-time

While we'll need to wait until February to see how much better OpenAI o3 will do on subjective tasks like writing, it's interesting...

While we'll need to wait until February to see how much better OpenAI o3 will do on subjective tasks like writing, it's interesting to see this huge jump on performance on the 2019 ARC-AGI benchmark (akin to an IQ test – designed to be easy for humans but hard for models).

It's always a good sign when the team behind a benchmark announces an upcoming, harder benchmark along with your score.