HN user
Low overhead engineering takes without the corporate fluff.
Altman may not be an altruist but the overall direction the organization takes is what matters in the end. And as long as that organization retains even a semblance of a non-profit (despite having a revenue centric approach to tackle competition in the field), it is still a preferable organization to Anthropic and Amodei.
This maneuver requires you to anticipate all the edge cases or error messages beforehand which is practically not possible in many situations. The moment something unanticipated happens or the model changes its processing logic, the tool call system stops working just like any other deterministic program or tool.
The capacity conundrum here is very unique. Similar to pareto optimization principle:
=> 80% of engineering problems == budget models like gpt-mini can handle.
=> 15% of complex reasoning, vibe coding == frontier models like Opus, still sustainable.
=> 5% of 'really complex enterprise grade reasoning' == need massive tokens, impossible at scale unless pockets are deep.
I'm a Senior Freelance Programmer, I can see many of my past and present clients moving towards the exact path you described. I keep warning them during meetings that Claude model isn't sustainable for long, eventually the VCs will come for their revenues and Claude will be forced to close their access to all but the most enterprisey ones with deep pockets. The mere electricity cost for that kind of high level reasoning and abstraction can't be subsidized forever. However, there are other forces which pull them towards Claude and AI workflows. Most of the clients are in a "wait and watch" mode right now, using LLM assistance for code generation but not fully depending on them.
Before LLMs came, there used to be the technical debt to deal with in a project, now there is also the added cognitive debt which is way more subtle and impactful long-term. If your source of truth isn't source code but a prompt (or even a series of prompts with branches) and the executor of prompts is a non-deterministic agent, I think you've already lost the battle there.
Might not be a very popular advice or even what you're looking for, but I think Anthropic banning you (or any user for that matter) is a blessing in disguise. The sheer amount of compute resources consumed by their high reasoning "pondering..." and "bloviating..." tokens isn't sustainable at scale. Eventually, they must ban everyone but those with deep and infinite pockets in order for this model to be sustainable and turn revenues.
Computers and LLMs are great at automation of low-level human cognitive tasks like memory, decisions and loops, etc. but struggle enormously with high cognitive tasks like reasoning, deep logic, nuance, etc. Not that it can't be done (Claude platform is proof that it can) - but the cost and scaling advantage in this realm belongs to the human brain, not the LLM.
Have you tried one of the Kimi K2 models or the latest GLM models by z.ai? The general consensus is that they're at least at par with Claude's class.
Great work, thanks for sharing.
At some point, there will come a saturation point for that "Opportunity cost FOMO train ride", and I think we are already past that point. Mythos class models are a whole different beasts and cutting edge on reasoning but not much use for the problem domains most developers are trying to solve.
The present Sonnet/Opus versions (~4.8) will likely be what everyone in the enterprise might end up using eventually. And even though local models aren't there yet, there are budget alternatives from the families of DeepSeek, Kimi, GPT, MiniMax, etc. available through APIs of NVidida, OpenRouter, Groq, etc. which are very much Sonnet grade.
It became a thing due to the experience of self-reflection. The million dollar question is how come humans (and few other organisms) are able to self reflect on their biology, life situations, logic, math and even consciousness itself? However complex and sophisticated a machine's brain is, be it biological or mechanical (AI/AGI), no known laws of science allows it to self-reflect. This is famously called "the hard problem of consciousness" in philosophy which remains unresolved to this day.
It was totally a rigged referendum, that bus hoarding propaganda somehow worked and the masses fell for the lies. But a great number didn't and it was barely lost by a few percentage points. David Cameron should have stayed and battled it out instead of resigning.
My bet is they'll keep subsidizing for a considerable period of time, at least 1-2 decades more.
Most AI companies are just testing the waters with paid tiers right now, their greatest fear with increased pricing is folks reverting back to wikipedia, stack-overflow and other public domain organic activity buzzing back to life; that will kill any RoI potential in LLMs forever. They're playing the wait game instead, observing how the digital sphere reacts to every little increase in price.
If that weren't the case, they'd be pricing at lucrative premiums already and even gotten away in short-term considering the increased dependency in the enterprise world. But that'd be like killing for the golden egg too soon and losing all long-term potential.
Once the folks are so addicted to LLMs that even writing a hello world program sounds like a nightmare and coming up with an article draft feels like reinventing Egyptian glyphs, that's when the real pricing hammer will come.
Is Vivaldi any good?
New prices are almost like double, a tad too high for an industry with wafer thin margins I think.
I've found some basic lifestyle changes to be way more effective at weight-loss and inducing well-being than a drastic chemical therapy like Ozempic. Imbibe at your own risk.
- Intermittent fasting (16x8 or at least stop eating after sun down)
- Daily morning pranayama (or 'breath work' if you prefer that term)
- Daily morning intake of copper utensil stored water (~8 hours or overnight)
- Weekly oil pulling for gum health.
This approach feels a bit stoic and neanderthal initially, less attuned to our modern and progressive society, but as effective as Ozempic in long-run (minus the side-effects or addiction).
If bots are being used to create chaos in society, it really isn't possible that the platforms themselves are just innocent bystanders here. It is technically possible and quite easy for the platforms to block bots if they really wanted to, in fact it's actually in their best interest to have human only organic activity as it increases the platform's credibility and reduces network cost. If they're still letting bots operate and actually post content on their platforms, they're likely in cahoots with the politicians.
I personally do not like the Javascript frontend ecosystem. It's hacks on top of hacks on top of hacks. But, do you know another way to deploy software that's cross-platform and basically free of gatekeepers?
One way is what I call the "Modular MVC pattern" that involves pure js routing and manual DOM manipulation without using any framework at all. You handle complexity in two ways: by modularizing the "controller" parts into multiple js modules for each route, and "view" parts into multiple HTML partials - and using the event bus pattern if your app gets too complex (as alternative to modern reactive frameworks like react/vue).
Shameless plug: I've tried to implement this exact pattern with limited success in Abhyasa Quiz App[1], a side project.
Not a tailwind geek myself but I think how they justify is "better to have a little extra spaghetti in your html code than create a truck load of spaghetti in your app.css stylesheet."
The alternative to using tailwind here is to define the specific style elements for each one in the css stylesheets yourself with something like this:
.bg-sky-500 { background-color: blue; }
Tailwind proponents argue that they avoid this "stylesheet hell" by picking ready pre-defined tailwind classes like bg-sky-500, etc. Plus they also argue that this workflow will increase productivity by standardizing "style mindsets" of your dev team who all will think "blue" means "sky-500" (for example).
Maybe it has use cases in deep or professional design work but for most backend or full-stack devs, bootstrap is definitely better than meddling with this structure.
A quiz app using minimal components like esbuild, bootstrap, etc. and pure js for routing and MVC: https://github.com/prahladyeri/abhyasa/
Also open-quiz-commons, the mcq dataset that powers the above quiz app: https://github.com/prahladyeri/open-quiz-commons
Expect petrol/gasoline prices to rise in coming future considering this new geopolitical risk.
Yes, most people in this age are trapped in the widespread phenomenon of "digital media induced dopamine traps". Your mind may never let you use old school slow information gathering tools (like books, manuals and software documentation) when easy and instant servings like LLM lookups and youtube vids are so accessible. Your mind might ask "why go through all that pain and trouble, what's the use of it?" and you'll have no easy answer to that.
Tunneling isn't that big of a toll on resource, it doesn't require storage/disk space nor compute power (CPU chips), all it needs is ingress/egress (spare bandwidth). A non-profit or decent business in telco can easily offer it, consider that many hosting companies offer entire package in free tier today (compute + disk + egress).
For several years, ngrok was practically free, only recently they've started monetizing once it gained popularity.
localtunnel[1] is one good option, at least for now.
Yes, similar thing happened with Ruby gems community some days ago. Did the contributors got their access back now or Shopify is still controlling all the affairs?
With no routes, no controllers, no business logic, how can the capability exist? These are the core components of a web app and require extensive coding. I know we might eventually get there but not with the present state of technology. There is something fundamental missing about "intelligence" which must be solved before AGI can be approached, throwing more money and nVidia chips at the problem can only take you so far.
Is your browser version slightly older? I've found that Cloudflare turnstile (captcha verification system) often fails unless your browser version is latest and cutting edge. However, this extra mile security is only for Cloudflare's own login page - Cloudflare verifications on other sites perform ok even on these older versions.
It's like putting all your eggs in this one huge basket called AGI.
It's almost like the only hope that the rich don't receive the Christmas gift is that AGI never arrives. AGI is literally that machine which will outperform even the best of engineers who then won't be needed at all.
Few companies have done it successfully like Red Hat, Odoo ERP and Sensio Labs (the company that builds Symfony framework).
The Linux kernel is a separate system layer here, it's the AOSP parts like the Dalvik Runtime (equivalent of JRE) and components built on top of it (such as Play Store) which are being subject to permissive licensing abuse. If AOSP itself was GPL licensed, it'd have been difficult for Google to create something closed like Play Store as it'd have been considered derivative work.
You're right that broadly speaking, there is very little that could be done to stop this but having a culture of "everything GPL" in an organization definitely helps. For example, Sun was farsighted enough, though they couldn't stop Oracle from acquiring MySql, Oracle was still forced to keep MySql under GPL and they were able to salvage MariaDB too.
Similar was the case with Java. Oracle tried everything in its power to control its use and direction including legal means, it's only thanks to GPL that alternative implementations like OpenJDK and Amazon Corretto still exist. We can't even imagine the state of these software today if Sun hadn't licensed them under GPL originally but used some other permissive license instead!
This situation would have been avoided if we, as community of engineers, had insisted on full and uncompromised open source (Stallmanist or GPL way) right from the start instead of going the ESR way of half-hearted open source where it's technically open but corporates get to have a free lunch and make abuses.
Like most coders, I also prefer the permissive MIT/Apache/BSD licensing for most software projects but incidents like these make me question the direction we are heading towards. They raise fundamental questions about freedom itself - looking at the broader picture, is having a restrictive kind of freedom (GPL) often more beneficial than having full permissive freedom (MIT/Apache)?