Compare both approaches to mature actor frameworks and they don’t seem to be breaking much ice. These kinds of supervisor trees and hierarchies aren’t new for actor based systems and they’re obvious applications of LLM agents working in concert.
The fact that Anthropic and OpenAI have been going on this long without such orchestration, considering the unavoidable issues of context windows and unreliable self-validation, without matching the basic system maturity you get from a default Akka installation shows us that these leading LLM providers (with more money, tokens, deals, access, and better employees than any of us), are learning in real time. Big chunks of the next gen hype machine wunder-agents are fully realizable with cron and basic actor based scripting. Deterministically, write once run forever, no subscription needed.
Kubernetes for agents is, speaking as a krappy kubernetes admin, not some leap, it’s how I’ve been wiring my local doom-coding agents together. I have a hypothesis that people at Google (who are pretty ok with kubernetes and maybe some LLM stuff), have been there for a minute too.
Good to see them building this out, excited to see whether LLM cluster failures multiply (like repeating bad photocopies), or nullify (“sorry Dave, but we’re not going to help build another Facebook, we’re not supposed to harm humanity and also PHP, so… no.”).
Orchestration definitely wasn't possible a year ago, the only tool that even produced decent results that far back was Aider, it wasn't fully agentic, and it didn't really shine until Gemini 2.5 03-25.
The truth is that people are doing experiments on most of this stuff, and a lot of them are even writing about it, but most of the time you don't see that writing (or the projects that get made) unless someone with an audience already (like Steve Yegge) makes it.
Not to mention data retention and upgrade management.
When an update script jacks up the guaranteed-to-be-robust vibed data setup in this first of a kind, one of a kind, singular installation… what then?
The pros have separate dev, test, QA, and prod environments. Immutable servers, NixOs, containers, git, and rollback options in orchestration frameworks. Why? Because uh-oh, oh-shit, say-what, no-you’re-kidding, oh-fuck, and oops are omnipresent.
MS Access was a great product with some scalability ceilings that took engineering to work past. MS Access solutions growing too big then imploding was a real concern that bit many departments. MS access was not dumping 15,000 LoC onto the laps of these non-developers and telling them they are hybrid spirit code warriors with next level hacking skills.
Ruby on Rails, Wordpress, SharePoint… there are legitimately better options out there for tiny-assed self-serving CRUD apps and cheap developer ecosystems. They’re not quite as fun, tho, and they don’t gas people up as well.
F# and C# are typed scripting languages. F# is quite similar to python in script form (.fsx), and has OCamls expressiveness, exhaustive pattern matching, and type inference. That results in highly expressive, terse, and ergonomic domain code.
A company I worked at years ago had a devious and exploitative approach to market domination: hiring older super experienced workers and plugging them into teams with young over-eager programmers…
People with deep industry knowledge who were trained up to be decent programmers (middling, but serious, consistent, and quality focused), setting the direction. Those domain experts were working with young dumbasses who would burn 60+ hour weeks to make sales deadlines and keep current with ever shifting platform tech that breaks all the time. SMEs baked into the core development loop, DDD-made-flesh essentially, with cheaper more junior devs supporting scale for less money and maximizing the SMEs vision/contributions.
It’s an obvious and effective strategy. I’d speculate the management skills it takes to setup are what keep it as a rarity.
For mature Enterprises my understanding is that the financial math works out such that the cloud becomes smart for market validation, before moving to cheaper long term solution once revenue is stable.
Scale up, prove the market and establish operations on the credit card, and if it doesn’t work the money moves onto more promising opportunities. If the operation is profitable you transition away from the too expensive cloud to increase profitability, and use the operations incoming revenue to pay for it (freeing up more money to chase more promising opportunities).
Personally I can’t imagine anything outside of a hybrid approach, if only to maintain power dynamics with suppliers on both sides. Price increases and forced changes can be met with instant redeployments off their services/stack, creating room for more substantive negotiations. When investments come in the form of saving time and money, it’s not hard to get everyone aligned.
Describing to Claude that I need an edit made in the second paragraph of the third section feels easy, comfortable, and straightforward. I’m using my speech centers, speech to text, and then I wait for a generation during which I hit my phone or Reddit. Poof, the text flies out like magic, taking 20+ seconds, then I re-re-re-read it to make sure the edit was good and nothing was lost in that edit. Oops, the edit inverted the logic of the paragraph, lemme repeat the above… and again… time flies! 2 hours gone in a flash.
Old and boring workflow:
I gruellingly move my mouse to open a file, then take a coffee break. I come back and left-click into the sentence that sucks. I hit Reddit to deal with the anxiety… I think, boo, and then type out the edit I needed. It’s bad, I fix. Coffee break. Squiggly red line from a misspelling? I fix. I google and find a better turn of phrase, copy and paste it in manually with a little edit. Ugh. This sucks. I suck, work sucks. Time sucks. 35 entire minutes of my life has been wasted… time to get another coffee and check Reddit.
———
Working with an LLM is kinda like working under stage hypnosis. The moment to moment feelings are deceptive and we humans are unreliable recorders of time-usage. Particularly when engaged in unwanted effort.
Google has had all this tech for a minute. Their restrained application and lack of 10x-vibe-chad talk make me think their output measurements are aligned with my measurements.
1 rabbit hole hallucination wrong-turn can eat up a lot, lot, lot of magic one-shotting.
> Working with an LLM is kinda like working under stage hypnosis.
Another post on HN likened it to gambling, in the way that slot machines work. Each time you prompt, you could hit the jackpot! But usually you end up with some mediocre or wrong, so you tweak that prompt and pull the lever again. It's and endless cycle.
I see Microsoft throwing spaghetti at the wall just in time as “AI” functionality hits government and educational procurement procedures.
The copilot product is obviously borked, and is outshone by ‘free’ competitors (Gemini, ChatGPT). But since the attributes and uses are so fuzzy, they have a minimum viable product to abort meaningful talk about competition while securing big contracts from governments and delivering dog water.
My anecdotal observations of copilot are people using competing products soon after trialling. Reports say Anthropics solution is in widespread use at Microsoft… a bunch of devs on MacBooks and iPhones using Claude to build and sell … not what they themselves use (since they are smart and have taste?).
I feel the market forces kinda point the other way, though, since the customization of the SaaS is also cheapening, but faster and more targeted than these internal teams. Over time I believe that’ll lead to more, not less, SaaS consolidation.
Let’s put the cost of code production at 0: regulatory compliance with payment processing laws or industry oversight is a recurring job that’s common for the whole industry when it changes. SaaS companies have hundreds of customers to attend, these become first class business functions. New demands won’t be in training data for LLMs, so someone needs to be doing this. SaaS has the funds and customer base to have dedicated experts at these functions, but it’s dead capital and nigh-impossible hiring in a tiny talent pool for the rest of the market… the delta to get Salesforce or SharePoint not to be total ass and fully customized is orders of magnitude smaller than detailing those foundations, and as people who sharecrop on platforms like those know, the devil is always in the details. Those internal teams just aren’t positioned to juggle both sides of that coin, they can’t be experts, mistakes can be existential, and the liability picture is so very ugly… coding is the least of it.
Into this, MBAs are not static. It’s not gonna take more than a few “vibe coding ate our CRM data” high profile snafus, or industry think pieces to map out why customization is faster/better/smarter, to get clear business dogma around this. A witty turn of phrase about focusing on your actual business.
I think ‘no one ever got fired for hiring IBM’ x 5 is on the horizon, and the evil marketers at Salesforce, MS, and the rest are gonna work hard to grow their piece of the pie. They have LLMs too, only with better models and unlimited tokens. And our executives will be checking directly with their LLMs about how to invest (the consultants, journalists, fanboys, and social media bots too…).
I remember hearing that people used it as a way to signal that they were too busy, too on the go, too important to use proper punctuation..it was an obnoxious c suite trend as long as I can remember. Like you're always trying to signal that you were doing all of your comms from your cell phone between meetings/travelling. Given this article's tone and content I would say that what the author is trying to emulate or convey , maybe subconciously.
Interesting. I am a millennial and I never did this, nor did I have any friends that did. But I know m nephews deliberately turn off the auto edit in there iphones.
The fact that Anthropic and OpenAI have been going on this long without such orchestration, considering the unavoidable issues of context windows and unreliable self-validation, without matching the basic system maturity you get from a default Akka installation shows us that these leading LLM providers (with more money, tokens, deals, access, and better employees than any of us), are learning in real time. Big chunks of the next gen hype machine wunder-agents are fully realizable with cron and basic actor based scripting. Deterministically, write once run forever, no subscription needed.
Kubernetes for agents is, speaking as a krappy kubernetes admin, not some leap, it’s how I’ve been wiring my local doom-coding agents together. I have a hypothesis that people at Google (who are pretty ok with kubernetes and maybe some LLM stuff), have been there for a minute too.
Good to see them building this out, excited to see whether LLM cluster failures multiply (like repeating bad photocopies), or nullify (“sorry Dave, but we’re not going to help build another Facebook, we’re not supposed to harm humanity and also PHP, so… no.”).
reply