siliconc0w

They'll buy your politicians who will give them zero checks on raising energy prices or poisoning your children's minds

Shameless self plug but also worried about the silent quality regressions, I started building a tool to track coding agent performance over time.. https://github.com/s1liconcow/repogauge

Here is a sample report that tries out the cheaper models + the newest Kimi2.6 model against the 5.4 'gold' testcases from the repo: https://repogauge.org/sample_report.

8M is kinda small potatoes - which is essentially the AWS business model. Sure you could build a cheaper thing but that is hard and this thing is right here, easy(ish) to use, and your company won't regret using it until you've moved into a different role.

We need to basically get rid of all exportable tokens. Require signing of all requests by a hardware backed key.

Working on a oss tool to help orgs identify where they can save on token costs: https://repogauge.org

Happy to run it on your repos for a free report: hi@repogauge.org

Definitely feeling this - the subsidized subscription plans are already starting to buckle.

The problem with AI is that it isn't like any previous technology. There may be temporary jobs to fill in the gaps but they won't be careers. The AI will do the process engineering and self optimization. The prompt witchcraft is a good example because today its totally unnecessary and doesn't actually increase performance, and they'll continue to make it easier to direct/steer the models.

We're literally trying to build an intelligence to replace us.

This is a pretty terrible decision and inconsistent with all sorts of all other standards. If I did legal research in Google docs, it'd be covered. If I went to a legal library and took notes, it'd be covered, etc

Yeah it does seem to be a new 'thing'. I'm betting it's like the heritage foundation who figured out age-restrictions give them a new end-run around the first amendment. Started with porn (because who is going to defend porn) and now they're going to slide down the slippery slope.

My experience is that Agentic Coding can legitimately get you mostly-working software. You do, however, still need to spend a few days groking, validating, and usually nudging/whacking it to conform to the shape you intended vs what the agent inferred.

It is pretty magical to go from brainstorming an idea in the evening, having ChatGPT Pro spit out a long list of beads to implement it, leaving it running over night in a totally empty repo and waking up to a mostly-implemented project.

My workflow uses a thorough design broken down into very specific tasks, agent mail, and a swarm of agents in a ralph loop to burn down tasks. Agents collaborate with mail pretty well and don't seem to need layers of supervision. If the tasks are well specified and your design is thought through, especially how to ensure the agents can self-validate - it seems to work pretty well.

I wrote an article on this if you're interested: https://x.com/siliconcow/status/2035373293893718117

I use rustfs (for local development, not scaled usage) and it seems solid.

I heard a CEO try to sell some tech they had as this hugely valuable IP and a competitive moat.

When I asked the team that is supposed to use it, "We haven't used that in six months and it wouldn't be missed if it were to go away".

+1 A huge amount of software - probably most - is not actually generating value and in many cases is actually reducing value.

I've seen teams build and re-build the same infrastructure over and over.

I saw a request that could have been met with a few SQL queries and a dashboard got turned into a huge endeavor that implements parts of an ETL, Configuration Management, CI/CD, and Ticketing system and is now in the critical path of all requests all because people didn't ask the right questions and the incentive in a large organization is to build a mini-empire to reign over.

That said, smart infrastructure investment absolutely can be a competitive advantage. Google's infrastructure is, IMO, a competitive advantage. The amount of vertical integration and scale is unparalleled.

Switched back to codex for the promotion. Opus at the start of the year was GOAT- just relentless at chewing through hard problems. Now it spins on pretty easy work (took three swings just to edit a ts file) and my session is like 1-3 prompts (downgraded to the $20 plan but still)

I'm excited for the AI lawsuits and litigation. It used to be squarely in the domain of the well financed but I can see the legal system absolutely inundated with AI generated legal slop.

I would love to see a stable test over time with a hold out set of easy/medium/hard challenges. I, like many others, have noticed a large drop in recent performance w/ Claude Opus (and Sonnet) and more sites like these would hold the labs more accountable to sneaky backend changes that nerf/degrade performance.

Maybe if you use the pouches but I rarely 'feel' like gum like I feel like a coffee.

If I'm on vacation I don't even think about having a gum. It's only when I'm working, notice I'm dragging, and remember the option.

We should at least ban the "preemptive" pardon if not all pardons. Pardon means forgiveness for a specific convicted crime, not a means to grant blanket immunity.

It needs to be a yes-and. We need to be better about fraud (digital ID + biometric proof on delivery of services) and we need a more equitable tax system that doesn't have the top .1% paying an effective tax rate lower than a school teacher.

Don't forget skyrocketing employer health care costs due to Government-endorsed monopoly and fraud that is the US Healthcare system. Robots don't need healthcare.

There is an argument that the VC-class is actually not a very good way to direct socially productive capital deployment and it'd be better to tax the LPs and direct the research using a public (but meritocratic) process.

For example China (not without its flaws) largely skipped over the digital drug phase and went straight into funding socially beneficial things like energy and materials production, AI and robotic enabled manufacturing, and large scale infrastructure build out. So ten years later they have a lot of socially productive things and we have a lot of socially unproductive things.

Nicotine is also a nice alternative to coffee. I still drink coffee but past late morning I'll usually reach for nicotine because the effect is much shorter.

It also has neuro-protective effects if you're an older gentleman.

I'm running out of Claude session limits in a single planning + implementation session even when using sonnet for the implementation. This isn't even super complex work - it was refactoring a data model, modifying templates/apis/services, etc. It has also gotten notably more 'lazy' like it updated the data model and not the template until I specifically pointed that out.

My backup has been Opencode + Kimi K2. It's definitely not as strong as even Sonnet but it's pretty fast and is serviceable for basic web app work like the above.

I'm working on a filesystem for agents for similar reasons - https://clawfs.dev - lmk if your team would like an invite..

As the demand for GPUs grows and supply cannot match it, the GPUs are going towards the enterprise and it'll be the haves vs the have-nots.

Instead of not driving to work to save fuel, frugal companies are going to have their engineers work on weekends to save tokens.

Do you have data for other benchmarks? +7% for HLE isn't nothing but it'd be more compelling if you could show you're consistently doing better with your method across more domains (especially coding, which seems like the primary use-case these days).

Quitting your job is a good first step but ideally you're supposed to sink $200/mo into tokens to code your AI-generated startup idea instead of hiring app developers.

I'm hoping a side effect of AI Slop in that by increasing volume it decreases value and people eventually start finding all Internet slop less compelling.

Local interference really needs $50k before you're out of "toy" capability.

Given how subsidized the subscription plans are and the oss benchmaxxing vs real life performance- you're paying much more for far worse models.