I can empathize with the author. I'm trying to get my first software business off the ground, and one of my moats was well, having written software. For 20 years. I know that writing software is a small part of a software business, but it was a differentiator. I'm still better off than a vibe-coder, but I think about that.
On the flip side, there's a story of an adventurous young couple who drove through the Democratic Republic of Congo [1]. They reported an incident where the road dipped and was submerged in water. Local men were standing around ready, for a fee, to chuck rocks in the water. The rocks piled up so the travelers' tires were able to grip the bottom and they were able to proceed on their journey. After they passed through, they observed in their rear view mirror the men now removing the rocks so the next driver would also have to pay.
I think wishing AI away to keep my moat is like wanting to remove those rocks. It's human to look out for #1 but it's a kind of tragedy of the commons.
I guess a lot of gravediggers lost their job at the end of the black death and no-one should be too sad about that. But I feel that the world will definitely lose something if LLMs put all the Indie software developers out of business.
Probably not an attack. The plane had a history of issues leading up to the crash. I think the saying "never attribute to malice what is attributable to incompetence" applies here.
The point of the article is to exonerate the pilots and expose a broader issue of honesty and transparency from Boeing and regulators that is covering up a systematic defect in Boeing aircraft design. The shared electrical bus renders other redundancies ineffective. The fact that the RAT has recently deployed 37 times uncommanded means another electrical failure at takeoff is a matter of time.
When you ask "why" 5 five times, it's never an individual. That's a big part of safety culture. Otherwise people cover their rears and don't divulge information that would save lives.
Aside: Isn't it wild that due to a software bug, Boeing directs airlines to reboot their 787-8s every 51 days?
Palmer's tweet said a similar thing and added, "we gave away 11.5 billion build minutes (~$184 million) to support OSS last year".
$184M in profit or cost?
It's called a loss leader, not a gift, and it's a marketing and adoption tactic. They already bought the machines which cost about as much to run idle as at 100% utilization. Might as well put that idle capex and opex to use.
Or just collectively bill OSS the $184M and stop signaling virtue.
We don’t know whether GitHub has idle capacity to spare. If they’re auto scaling their resources like a well-architected workload should, then they probably don’t. Even if they’re not downscaling their compute, they may prefer to run other low-priority jobs on it.
> Self-hosted runners: You will be charged for using the GitHub Actions cloud platform from March 1, 2026
The GitHub encrapification finally affects me. I am militantly unwilling to pay per minute to use my own computer. Time to leave. I can trigger and monitor builds myself thank you very much.
To help draw out some concrete software requirements, can you expand on what a "good way to manage the failures" would be and how what you've tried "does not work very well"?
- How many failures per day do you experience?
- How long do you need to retain records?
- What's the lifecycle of a failure? How does it get recorded, who investigates/triages, who responds?
- What are your custom scripts doing? Importing/exporting tickets?
- SSO, audit, compliance needs?
- You said JIRA was slow and limited. Is that just the UI or is creating/managing tickets too cumbersome? (Or yes to both X-D)
- What specifically broke with Google Sheets? Not enough rows/too slow?
You may well be looking at custom software since a lot of the apps that come to mind are either focused on aggregation (Datadog, etc) or high-touch tickets e.g. product development or customer support.
The process is pretty informal, so there are no hard requirements. that said:
- Failures per day: let's say 0-100
- How long to retain records: no hard requirements? I guess a few months at least, some failures are pretty rare
- What's the lifecycle of a failure? Scripts record it, team members investigate it and assign to "root cause".
- Custom scripts:
(1) create ticket per failure
(2) create failure reports (to prioritize work - for example if there were 50 failure reports with root cause of 'github was down' , the priority of "set up github mirror" will get bumped up)
(3) mass-update tickets (for example if github.com is down, there will be few dozens of failed processes because of that)
(4) handle rules for automatic classification (again, if github.com is down, it'd be lovely if I can have a rule: "for the next 48 hours, every ticket which mentions github.com and 503 is auto-assigned to 'github was down' root cause")
- SSO, audit, compliance: nice but not required
- JIRA problems: search sucks. "Find similar ticket" sucks. Rules are missing (or need admin). Even something as simple as "close those 20 tickets and link them all to ABC-1234" is impossible.
- Google sheets: not enough automation. At least I can do "filter rows, copy-paste the 'root cause' field into all of them", and it is pretty fast, but: multi-line outputs don't look good and there are no automation (we did not explore App Script, maybe we should have...)
And yeah, I am getting the feeling this would be a custom job. We have resources in house to do so, but I was hoping there was an existing product. Surely there are people out there who run batch-like jobs and want them to be reliable? Something like data conversion jobs, CI builds, training jobs, etc...
Perhaps it's a good thing for generative AI, I've heard it's pretty good at making websites (and security/availability is not an issue, as this will be internal website not exposed to internet). Or I may revisit Google's App Script...
Thanks for your reply. I suggest looking at Airtable and _maybe_ Linear. They have API and automations. You could likely get AI to rewrite your scripts.
If those don't work, you may have a business case for building it.
I'm a founder and dev looking to for a good problem to solve. If the need could be proven (e.g. 10 people with decision power said they wanted it), I'd consider making it.
https://www.slater.dev/2025/08/llms-are-not-bicycles-for-the...
reply