Dane Albaugh

Everyone talks about AI in manufacturing like it will be easy to plug and play. Manufacturing is a hellscape of legacy software infrastructure and operating practices that have far outlived their utility. There is a lot of work to be done on the API layer

@CalebChamberla6 Raw material extraction (mining ore) and processing seems much more likely to be the bottleneck in that future? I think you are directionally right though

Anyone else think the SCS raise is actually an AI bet? If AGI becomes a thing and machines can design, manufacturing becomes the bottleneck. Scalable manufacturing and existing capacity will be critical. Enter businesses like OSH and SCS, designed for high mix scale from first principles. I'm not weighing in on that likelihood, per se. But the multiples are wild, and interest from Sequoia & co. is historically aberrant.

@DenverRayburn lol. Great stories always do

that last thread really brought out the crazies

Finally, seeing some tomatoes and looks like our drought is gonna be relieved next week by some rain

The issue with on prem in older factories is not that on prem is inherently bad. The problem is that proper redundancy costs real money, and leadership often refuses to fund it appropriately In theory, you can build highly available on prem systems. In reality, most factories are running decades of accumulated vendor software with bizarre licensing restrictions and hard dependencies A lot of industrial software refuses to run active-active or even on multiple machines because vendors license per server and do not want customers duplicating instances. So even if engineering wants redundancy, the software itself creates unavoidable single points of failure The result is a spaghetti stack of legacy Windows servers, vendor specific VMs, proprietary PLC tooling, ancient databases, accounting software, and custom integrations all glued together over 20 years Higher ups usually see redundancy as “extra cost” instead of production insurance, so the infrastructure ends up fragile by design Sure, when you build your factory from scratch I’m sure on prem will make sense but that is because you will invest in it and know what you are doing. Most factory owners are clueless but are still obsessive about on prem. My whole point is this needs to be thought of at a system level and all points of failure should be considered and redundancies added. I can do that just fine on the cloud

@dane_albaugh Brother… I like you, but the fact you had an outage with onprem might be the skill issue.

I have had an on prem server take down production before. I have had an internet outage disrupt production before. AWS has *never* resulted in an outage in our factory The conviction that on prem is bulletproof and “cloud” = bad in the factory is more often than not poorly concealed skill issues

@aarondfrancis Cool!

@dane_albaugh yeah I did the original design! now it's just finding stuff like "I tried to update the content without adding a name and it failed" and it's like... oh yeah, that's probably needed

I'm building an API that will be primarily used by agents. Each time the lead agent makes a change, it spawns a new agent, tells it use the API to do a set of tasks, and report any papercuts. Then the lead agent fixes the papercuts, spawns another agent, and tries again. ♻️♻️

Cucumbers are coming along nicely

Continued on our section hike of the AT this spring. A good dog is a tired dog

The @PlanetScale MCP is really wonderful. It makes finding and fixing slow queries so easy

@joshpuckett Wow. You have convinced me - I have to see this in person!

If you are a fan of faith, architecture, or even just beauty itself, you must visit la Sagrada Família in the afternoon light. An absolute masterpiece.

@TheDarkGoldMan Cucumbers, tomatillos, jalapeños, habaneros, bell peppers, zucchini, squash, pumpkin, collard greens, broccoli, and lot of varieties of tomatoes. Also, some minor stuff e.g. chamomile, flowers, etc.

@dane_albaugh The best boomer system! Haha! What are you growing, the old fashioned way?

Amateur system

Now, any object ID in our request logs are clickable links that allow you to navigate directly to the object in question. It makes it a lot easier to debug requests

@CalebChamberla6 @easypost Might be worth having duplicate integrations with a separate API for failures like this. I have that setup in Augno so if our shipping aggregator goes down we can still ship, even with some reduced functionality

EasyPost is down. Getting server error responses from the API. But their status page says everything is fine. @easypost , what gives?

banger articles you should read

99.8% of bun’s pre-existing test suite passes on Linux x64 glibc in the rust rewrite

@barbinbrad Cool feature

Item rules. The UI needs polis but you get the idea

@CalebChamberla6 At what age did you start OSH Cut?

The first 20 to 30 years of your career is just practice.

Fun! Although it is obviously keyed in on Stripe conventions (e.g. using POST for resource updates rather than PATCH, using POST for all custom actions rather than using PUT depending on idempotency guarantees) there were some interesting insights. It does seem to hallucinate some issues. For example, it suggested resource types on objects were not present, but our API returns these with every object (with minor exceptions). docs.augno.com/api-reference/… It also suggested we call our endpoint "Revoke" rather than "Delete" for the api keys, but we already do that. docs.augno.com/api-reference/… The criteria for "Custom methods correctly use POST with verb-prefixed names" was flagged as passing when above it flagged "PUT used for custom action methods" as a critical error.

Stripe obsesses over API design. For Stripe Sessions we turned our well-worn API design principles into an app that reviews your API for $2. We liked it so much we’re making it public for 30 days. api-reviews-by-stripe.vercel.app