Not quite the same scenario, but it's already plausible to have a situation where every subagent is allowed to spawn multiple subagents, in which case we'd have literally exponential credit consumption growth...
"i have to burn $10k in tokens to meet my end-of-month work quota. spawn ten sub-agents each of which is allowed to spawn as many sub-agents as it likes to create an analysis of the code in these files based on the precepts of the 13th century German philosopher Noodleheinz".
Startup working on AI Healthcare solutions (plus some random experiments) | REMOTE(US, Pacific time zone) | Full time and project based
Full time engineering role (Remote, US person required; Pacific time zone)
Stack: C#, Python, Gemini API on Windows, lots of legacy systems/constrained enterprise environment. Health systems experience preferred. Understanding of cost effective, reliable LLM API use. Should be flexible: do PM-like work with less-technical client domain experts, solve problems end to end, willing to data cleaning and detailed work for client goals and security/compliance gruntwork, even when it's not technically elegant. Self motivated, willing to understand legacy and undocumented codebases, but also willing to take instructions.
Ideal candidate has real (pre-Claude Code) experience, but is ready to do fast iteration in post-AI world, while actually reading what the AI does. We're ok if your stack is different, if you're both technically mature and up to date on modern fast-turnaround AI-based development. Probably in the 5+ year experience range.
REMOTE anywhere, project based or initially part time: startup generalist hacker, maintain systems and websites, code one-off prototypes. Should be good enough to code without an AI in a few languages, but an effective AI harness coder. Can do a small but complete app/project end to end, including basic marketing. 2+ year exp or a new grad with a good portfolio or major project; happy to consider more experienced person. Good role for someone who's working on their own startup to pick up some side money and experience. Projects could be anything from game-like apps/novelties to voice tools to utility apps.
REMOTE anywhere, project based or initially part time: help on some more experimental mathy projects. M.S./Ph.D. in statistics, physics or related, with pragmatic programming experience. Possible projects: gaussian process modeling; designing evaluation metrics; compression; Bayesian tools for radiology imaging. May be good for a grad student part time. This is the fun stuff the founder would rather be doing himself if he had the time.
REMOTE (US person) Half-time remote admin/chief of staff type: deal with bookkeeping, compliance, communications, social media, phone calls - save us time on everything non-technical. Should be effective AI user; should have some relevant experience supporting tech business.
Please reply with a salary target. may2026-jobpost@waldrews.anonaddy.com
File systems are nice if you need to do manual or transparent script-based manipulations. Like 'oh hey, I just want to duplicate this entry and hand-modify it, and put these others in an archive.' Or use your OS's access control and network sharing easily with heterogeneous tools accessing the data from multiple machines. Or if you've got a lot of large blobs that aren't going to get modified in place.
What the world needs is a hybrid - database ACID/transaction semantics with the ability to cd/mv/cp file-like objects.
The new workflow will be "AI, I need to view this text file and add some words to it. Create an app that displays it in a scrollable window, respecting the encoding. Now move the cursor to the line below the three dashes... no, the other three dashes..."
well, sure, that uses a large number of processing cycles for each small operation. But asking a frontier LLM to evaluate a lisp expression is more or less on the same scale (interesting empirical question whether it's more or less). And, if we count operations at the brain neuron level it would take to evaluate one mentally....
And they're pretty much the only example of an embedded browser architecture actually performing tolerably and integrating well with the native environment.
That's one maxed out RAM configuration. Back in my day, we had 4k RAM, about 3500 bytes usable from BASIC, and that was enough, unless you were rich enough to have a 3k memory expansion cartridge. But really, if you need that extra 3k, you're just not writing code efficiently enough, right.
Back in the 80s, I was lucky that my father was an electronics design engineer, so he built a 24K expansion cartridge for us. I agree that there were some great games for the unexpanded VIC 20 though, such as Rockman. I loved that game. So many levels for a small game.
I had a home brewed ram expansion board (still do actually, in a box somewhere...) I powered everything up a couple years ago when my kids found it and asked what the heck it was. Still works
My original VIC 20 machine that I had in the 80s still works as well, but a few things have been replaced along the way. I still have the same 24K expansion cart that my Dad built 40 years ago and it also still works.
We're just not going to see any code written entirely without AI except in specialist niches, just as we don't see handwritten assembly and binaries. So the disclosure part is going to become boilerplate.
In the old era, the combination 'it works' + 'it uses a sophisticated language' + 'it integrates with a complex codebase' implied that this was an intentional effort by someone who knew what they were doing, and therefore probably safe to commit.
We can no longer make that social assumption. So then, what can we rely on to signal 'this was thoroughly supervised and reviewed and understood and tested?' That's going to be hard and subjective.
Personal reputations and track records are pedigrees and brands are going to become more important in the industry; and the meritocratic 'code talks no matter where you came from' ethos is at risk.
The water from sewage might end up there after it's extracted and sanitized, but all the solids have to be disposed of too. Those solids, plus the leftover chemicals used to extract and sanitize the water, go to landfill.
