brandensilva

The cycle's will never gonna end. It's just gonna be incredibly faster now.

Yeah, I'm more interested in them covering the majority of people's needs and use cases than I guess having a hundred percent compatibility with libraries I'll never use.

I'd much rather it be a focus that I could run bun in serverless, sandbox, wasm, or workerd-like environments for agentic workflows where bun is the runtime to power additional functionality for AI agents.

I assume this is why Anthropic purchased them beyond just a PR stunt and where the project is going.

If it was one memory leak, it would be meh in the fix list, but the fact they've announced that they've eliminated all memory leaks because of Rust in about eleven days that's what's wild to me.

We are living in a new world when it comes in how software is prompted into existence.

Elixir started adding a type system as well after Typescript proved to greatly help AI feel confident in its output.

So yeah others are catching on the importance of a type system to enforce that successful contract for AI agents.

But there will be and his point is they cut off supply to make room for the new M5 ultra which I hope has 768GB or more of memory.

So reckless but needed.

It's easy to see why people fear AI when our leaders talk of a future where many are jobless and replaced with no solutions to fill in the gaps.

AI adoption is a leadership failure more than a tech one right now. If you make people feel empowered with it, it can liberate work-free lives that humanity benefits from. If you use it to destroy people's livelihoods with no options it's not going to survive a revolution.

The irony here is the red Republican states are the ones voting for this who will be most impacted by it.

Thoughts on a M5 Ultra 768GB if it drops? What's the price to make it worth it for you over a spark cluster?

I'm wanting to run Kimi 2.6/2.7 GGUF on it and just slap it in the server rack, but trying to decide if a spark cluster makes more sense.

This is correct, they are subsidized but it's the training cost that costs the most with a majority of people hitting cache for most queries for inference.

Devs are hard to market and sell too I've heard. It's likely because they can build a lot of the stuff out there themselves when pressed. They have the most app exposure so are opinionated. It's why most devs take the open source spoils while everyone else avoids GitHub in general. Although AI has made it easy to setup locally, many still don't see the value of controlling their software or ai agents fully like devs.

Yeah the cost doesn't justify the value. Spending x3 on tokens is meaningless if they aren't seeing the profit side scale with it. Going faster on product features doesn't mean it translates to more money. In fact I'm not even sure some customers can handle the speed at which a team could move with feature delivery.

So scaling horizontally in different markets seems like an advantage if a product is already mature. Which is exactly what Anthropic and Open AI are doing because they want to put their tentacles in everything.

That's exactly what I'm doing. Push for model flexibility, cost control, and reliability.

The moat is to be anti lock in. The open source model needs to be amped up on steroids to avoid this permiating every facet of your org you cannot escape.

Exactly, it's far bigger than Meta when the government's are pushing a larger agenda here.

The assumption is you have to control people to enforce laws. They keep pushing this notion that is a requirement to keep people safe. That somehow if we have big brother AI surveillance everyone will be on their best behavior.

Oracle, Palantir, Meta, and other mega billionaires push this agenda because who is going to stop them from controlling society and getting absurdly powerful and wealthy from it?

I find it strange for these people to accept such a defeatist attitude because I'm the opposite.

I mean I will just not use the service and I'll seek out alternatives that are open source or create my own. I'll do anything possible until I'm the last one standing if that's what it comes down too.

I tried to sign up to Telnyx and they had the same crap from an unreliable data-breach and being-litigated persona identifier. I passed on that.

I've already been going down this road as I've abandoned Google and some of the big cloud providers in favor of smaller companies who aren't pushing these policies.

It isn't hard to click cancel. It's just people favor convenience over their own freedom because they have never experienced not having freedom like our founder's did 250 years ago. The problem is once freedom is gone, getting it back requires blood spilled and political reforms and revolutions based on what history teaches us.

Illegal tariffs, executive usurping congress power of the purse, Noem funding herself and friends with a commercial from an unknown entity with tax payer money, people in ICE/FBI handing over undisclosed unaccounted money in brown bags, insider trading is rampant, using funds inappropriately to fly girlfriend places that isn't official business, illegally using private money to fund public projects, taking bribes from foreign nations like jets and such violating emulation clauses, passing no bid contracts to people you know, using the pardon power inappropriately to pardon crypto scammers and other white collar crimes, moving notorious Epstein related criminals to a low security prison without going through the courts, avoiding justice for sex crimes of the rich, using the DOJ as a political cudgel, and the list goes on.

This article would do better to show good use cases to draw strengths to their premise but I agree with your take that frames matter less than the total feel of the transition. Some of them definitely could be improved of what he outlined.

I find AI to feel real nice for pushing delight like this further than I used to have time for as it was never a priority.

I've been wondering why they attack science outside of they think it is woke and liberal.

It makes no sense to cut off the hand that saves you even as a rich billionaire who wants to control people in a fascist society.

I agree, leave the training to open source federations that roll out like operating systems. Minimal training over time.

Then have inference go down to the next layer to use those models as a P2P decentralized network.

Maybe like open router could tap federation networks.

It is definitely the fun part to me and scaling a factory is the challenge we all decide every day we write code.

I think of AI agents as a factory unlock too.

Anything of quality needs inspection, review, and so on before you ship it. The human in the loop step is critical for value delivery.

Many software companies don't think of themselves as factories but they ship a product to a million customers and it solves a work unit of value for each customer.

Now AI agents may eventually reach a similar potential where many kinds of work can be manufactured by an agent.

The difference I think we are saying as engineers is shipping code isn't the unit of value if you want to turn a code agent into a code factory. It's just a byproduct of the value. Code is the liability, tool, or contract to delivering the value. Poor engineering results in poor output that cannot scale or poor quality that no one wants.

Without us inspecting and reviewing the output you risk the value not being delivered. Also without us, in the past you don't have a factory to begin (although vibe coding has collapsed how soon people can setup a software factory now), scaling it takes engineering effort. We build the factory and also ensure it is operating well to deliver that value to customers.

The ability for a program for loop to do a million iterations is foundational to our knowledge. AI agents just scale that up and is one of the tools we use to get there.

I'm not scared of agents making code cheap. I just see them as another tool in the arsenal to build scalable systems now. Yes some people don't need to scale and some don't need anything of quality. That's fine and is welcomed to improving people's lives. If they want to scale or need a quality factory line they come to us to deliver that still.

I think it's myopic to think this isn't coming. But it's also short sighted to assume our value is gone at building, shipping, and operating factories.

We blocked YouTube recently in the household for all devices but one approved tv device that our kids are only able to watch with us.

I let my oldest daughter at 10 watch stuff there a couple times a week which she largely watches Minecraft videos. I know everything she consumes for now.

Eventually that will stop and she's on her own when she is more responsible as an older teenager but the important point here is this isn't helicopter parenting, it's survival at protecting her brain from dopamine overload making her a content addict.

I don't want to go full Amish as I think it's important to prepare our kids for the inevitable world they will be exposed too but I feel I'd fail them letting them loose.

Micron is building a factory near me but it's years out.

The only hope here is China floods the market which would be a good or bad thing however you look at it.

I have found this to be the case as well. As developers we are just really good stewards of the code because we obviously have knowledge to make sure that the code is engineered in a way that it can scale and grow without tech debt becoming unwieldy.

I found AI to be pretty bad with like a bare bones code base without solid patterns in place already. It works but it's just monolithic files galore. use effects hooks everywhere. Nasty state situations with poor data practices. Security vulnerabilities up the wazoo.

It's weird to have this conversion with them. Like yeah your code works but it's so tangled up it's hard to reason about where to start to begin to unwind it all sometimes.

It can be done but cleaning up someone else's slop is the exact reason why I hate AI. It was hard enough to review great code and be critical, honest, and fair but we knew it was an essential part of the process, helped build shared understanding, and was a way to learn from one another.

Whereas throwing in jumbled garbage to review just feels like a waste of our brain cells we spent decades earning by embracing the craft.

Eventually infrastructure will be more simple to orchestrate too without faults I suspect from well developed devops harnesses. The risk and scale companies are willing to accept will still fall on humans for some time even then. I don't see most people vibe coding a million user app that has deeper needs than the basics we see now.

Yes, exactly. Any engineer deep on this stuff right now understands that grounded predictive engine sprinkled with RL training and are discovering what that means in terms of its strengths and weaknesses for company use.

China is about to flood the market and prove this notion wrong. If there is demand they want to meet it with supply.

But to your point, that is exactly how American companies like to play now. No one is stopping them from screwing over the consumer.

I have a Micron near me and they are building another chip facility but we are years away still so I suspect China will beat them to the punch.

Right more simply put it's great at being a copy cat, exploring similar data points that match your token needs.

It is not great at decision making or judgment calls that don't have a well defined spec or plan in place yet; like unofficial or unapproved tokens if you will. A lot of this stuff simply never has had specs as it has been internal to how companies work and their secret sauce.

The closest thing we have are governance and compliance policies due to legal/business needs requiring it so it's far more well documented than operational ones in how we work. It is more about the how versus the what here I guess is what I'm saying.

But yeah this is why it does great when there are tests, design systems, evals, and other artifacts to mirror. Far more reckless and unpredictable without these things, but still great for exploration and finding the data output you seek.

Cleaning up tech debt is a must with AI for a lot of these orphaned utilities and anti patterns.

My hope is eventually open source models get far enough along we just train the models on company specific code needs.

Until then it is a lot of wack a mole with skills, hooks, system prompts, and interweaving deterministic needs.

If that is fraud then company evaluations are fraud too. Case in point SpaceX and it's smorgasbord of other companies rolled into it to save them.

Who protects the consumer when they have been gutted of any power?

I remember being that kid in high school who ran math and logical problems hard which contributed to me being very technical and to learn to push through painful mental challenges on the regular. Out of most of my graduating class there were not many of us that went on to become engineers for a reason because it isn't easy work by any means and I'm guessing is quite draining for people who don't use their brain like we do.

So while AI will change the industry I don't see any reputable company firing the smartest ones in the room for junior level intelligence.

Even with it advancing someone has to be responsible for when it screws up which we know it will.