Allen Pike

Allen Pike

@allenpike.com

Building usepowerplant.com, hosting Infer and It Shipped That Way, and writing about software and other tricky things. ⭐️ Vancouver • @apike elsewhere • allenpike.com

880 followers217 following167 posts

Telling agents to fix mistakes they made isn’t satisfying. Having them help ensure future agents won’t make the same mistakes is much moreso

Scaling coding agents across teams requires strong guidelines (what does good code look like) and guardrails (rejecting bad code). Fast system-1 models like Jev are a big opportunity for faster, cheaper, and more plentiful guardrails.

If you want trains that can run every 75 seconds, all you need to do is automate! Otherwise you get merge conflicts, which are annoying for code, but worse for trains

YouTube's parental controls have, for now, deterred our children from getting up at 6am to watch brainrot on the TV. Feature request: A checkbox to block all content that samples "Industrial Samples & Impacts by Bluezone" aka the Vine Boom. Also screaming

Cool thing about letting agents see the prompts that created the work: they can help fix recurring mistakes (e.g. over-verbose UI copy) by checking how humans are reprimanding the bots. All three of us had been pushing back on this mistake, but hadn't yet improved our shared AGENTS file to fix it.

With macOS 27, we’ve refined the experience from top to bottom, while still making sure to keep classic experiences that users have come to expect from their Mac

As you create more projects to manage your agents, you might find yourself quickly kicking off tasks in different projects. Thankfully ChatGPT née Codex has a keyboard shortcut for this frequently-used switcher – invoked using just two hands and, optionally, an eyeroll

Claude Fable is useful as a thinking tool for decisions. But I’m going to be honest here, because you deserve the load-bearing quiet part said out loud: it tries too hard to sound deep.

Now, I’m sure these new butterfly refreshers are delicious. I’m gonna try them like everybody else. But I’m not sure how I feel about Starbucks selling drinks made from insects

I should have known that 20 minutes early was *not* early enough to line up for Paul Bakaus’ Impeccable skills workshop at AIE World’s Fair

It's fun to talk to an agent and have it respond with rich visuals – but building low-latency experiences like this is a pain in the ass. In my AI Engineering World's Fair talk, I share some of the joy and suffering from our work on this problem. m.youtube.com/watch?v=65X0...

m.youtube.comVoice In, Visuals Out: The Agony and the Ecstasy - Allen Pike, Forestwalk LabsThe latest AI models have made it possible to go past text-based chats, and build what Andrej Karpathy argues is the pinnacle of AI UX: voice-in, visuals-out. In this talk, Forestwalk Labs…

It’s cool that they made a thumbnail for the video, but a shame that they’ve blocked the majestic SVG penguin with my big head

“Stop! Who would refresh the Flight Status of Death must swipe me with fingers three, ere the other side he see.” The United app is less bad than most, but there’s a reason people love Flighty

I don't use Safari often, but after downloading an invoice from my cell carrier, this delightfully ancient fossil popped up out of nowhere. I couldn't quite bring myself to "Allow" and see what account Safari thought it'd found

A clip from this morning's daily huddle. It's always fun when your own tool does something you didn't realize it could do! (I'll allow myself to feel proud for one day. I'll resume insisting it needs to be 10x better tomorrow.)

I'm busy coding and don't have time for macOS' whining about RAM! Can somebody who's good with computers help me figure out which app is causing problems here

It's annoying when you're in a meeting, and somebody describes a bug really well. So you're like, "We should file an issue with what you just said" – but then have to actually type all that stuff out. Like an animal. So now Cedarloop will do it for you.

Today's Claude Design launch addresses the fact that chat loops suck for iterating, exploring, and polishing visual spaces by adding first-class support for knobs, dials, and other exploration controls. Goodbye "try making the cities bigger, maybe 30% bigger?", hello generative knobs and sliders. ⭐️

When Apple Intelligence first rolled out in 2024, I had to turn off the Priority Notifications feature because of poor reasoning. But trying it now, after all the advancements we’ve seen since then, you can see they’ve really dialed in what is urgent

As systems include more layers of AI, it gets difficult to intuit why they behave as they do. For example, why is one of the Google results for “allen pike" the Wikipedia page for director Spike Jonze? Zero people searching for Allen Pike are looking for info on Spike Jonze!

- But Google knows "allen pike" is a "person" search. - And "allen pike" has a similar embedding to the username "apike". - And ”apike" has a similar embedding, as a typo, to "spike". So, Google relates "allen pike” to “Spike Jonze". And that's why AI devs need good tracing and analysis tooling 🫠

Despite the challenges, it's exciting to imagine that 2026 could be the year where AI becomes able to complete multi-day tasks, perform useful household robotics work, or even produce working CSS layouts where only part of the screen scrolls

One fact about the Pacific Northwest is that half of its software talent is in the cities, and the other half live on random-ass islands you've probably never heard of. The longer you live here, the more islands you'll learn that some talented developer is currently working from.

TIL there's a cracked Node dev working from Camano Island (number 10 for me, so far). I feel like the different islands should have rarities and different points. For example obviously Vancouver Island is only 1 point, but if you can collect Galiano or Guemes or something, that should be worth more

It's a common pattern to make users type "DELETE", or even the name of what they're deleting, for irreversible delete actions. This forces them to think before potentially deleting something critical. Intercom takes this to a new level, forcing you to consider: "Should I be the one deleting this?"