I'd be tempted to try something similar for this season. But this is what I generally pay HR Block for; they should be relatively vetted (granted i don't know this for a fact).
like do you have tax accounting experience or do you intend to have someone who does help validate test returns to make sure they look right?
1) I've built the engine and tested it based on CPA or externally reviewed real returns. You can see that in the scenario tests
2) I also have a mechanism for people to test the engine against their previous returns privately and then send bug reports of something wrong
The engine currently only works for 2025 because the 2026 documents aren't out yet from the irs.
There is also definitely a place as well in the tax world for general estimates. For example calculating the tax consequences of a stock sale or iso exercise can get really complicated without a system and cost $800+ with a CPA to do.
Was it in the "FF 5"'s opening where they carjacked a tanker with ~$50k worth of diesel? Nowadays a 90 m³ trailer full of DDR5 RDIMMs would be what... $200M? $500M?
I've been doing some research lately on my own (aside from the major benchmarking going on) and I've definitely found cases where medium models do much poorly and cost more than the higher models. But on the other hand, if I can give the medium models certain skills or other resources to complete a task then they perform better in cost than the higher model (which, given the same resources, sometimes gets more confused than the medium does).
At least for my use case of evaluations of a set of features.
You’re talking about something else (reasoning levels). The parent comment is referring to e.g. GPT Terra as opposed to Luna, Sol or Astra, Claude Sonnet as opposed to Haiku, Opus or Fable, or Gemini Flash/Hydrogen as opposed to Flash-Lite, Argon or Carbon.
Oh, good catch. I do use Sol in production, for cases where I need quite good results, but can't afford to run in production the highest tier model due to pricing (so it's like the highest tier model I can afford for that use-case).
To be more specific, this is for a support automation, where the model has to be quite accurate in following rules and generate responses, and Luna models or lower-tier models often make bad mistakes and generate wrong responses, so are not really usable for sending data directly to the user.
I first learned to program in middle school with Logo, and the extent of that was "telling the commander" (or whatever the term was) what to do in almost plain english.
That's one of the things that got me into learning to write code in the first place.
As long as the child in question gets to explore their creativity like this and it instills that sense of learning and diving deep then I'm not sure there's a problem here.
When I was a kid I got into programming to make games and websites partly because of the challenge of it. It was quite difficult to do and whenever you made something work it felt very rewarding.
Well now why would anybody bother writing that code by hand? Perception will/has shifted very quickly to coding being something that is simply done by AI with human oversight.
If I was a kid now maybe I would be using AI to create games or something but I think I would probably just get bored quickly and never bother to learn coding.
Sure, but I have gone from enthusiastically recommending a career in programming to being very cautiously optimistic that the strong kids will still find a place.
Not learning to code might be ok; maybe they can learn to fix John Deere tractors instead.
> very cautiously optimistic that the strong kids will still find a place.
Well yeah I mean that's always going to hold true if we are just talking about things in general
> maybe they can learn to fix John Deere tractors instead.
Yes these roles will probably shift to hardware more and more. But it's very different, much harder/impossible to teach yourself to the same level as you can with coding. The knowledge I gained staying up until 2am writing Flash games transferred directly to my career in software engineering
I haven't used an AGENTS or CLAUDE file as much lately. I'm not sure how much more it helps, but generally I try to have one or more project md files. These are curated to whatever I am currently working on. Perhaps references to certain modules, commands, schemas for DB, etc.
This way the agent doesn't get too much context it doesn't necessarily need.
Again, not sure how much more effective or efficient it is but it seems to do better for my needs at least.
Not too dissimilar from my approach at work these days, sans the "up to 32 agents" bit.
I try to work on planning docs, scrutinize them a fair bit to look for edge cases or anything that stands out as a problem, whatever.
Then break down the work further into meaningful chunks and hand those out to agents. Mostly one agent at a time but with the directive to use subagents as needed for implementation/review/whatever other perspective I want.
I feel like I'm exercising my brain similarly to how I had in the past. Before LLMs we'd whiteboard things similarly and figure out if a design made sense and like you said, chunk it into Jira tickets. it's really just the implementor (?) that's changed.
Just the other day my teammate approved my PR and despite the repo having that merge when ready feature enabled, it kept not working and throwing some random error.
IIRC we unapproved it, approved without that enabled, and then manually merged it.
their humanoid robots really creep me out. I've been to a few conferences for work over the past few months and they usually have a booth and then have someone walking them around by remote control.
The conference venues have concrete flooring, so I just suddenly hear loud rhythmic thumping as these things stomp around a corner.
After the initial "huh that's a cool robot thing" thought, it really started giving me the heebeejeebees.
Interesting theory. It implies existence of non-human humanoids in our distant evolutionary past where recognizing them would give us survival advantage.
There would have been plenty of those - we keep finding near ancestors, neanderthalis, the hobbit one. Some say the Monkey King from chinese legend is a folk tale of a close species
Also because they make them all faceless and sleek and alien, like a human shaped iPhone. If they instead looked like little Gundams I think they'd be cool.
Multiple non-developer colleagues at my workplace are trying out all kinds of MCP connectors in their Claude instances. Maybe when MCP was mostly relegated to coding harnesses it was more of a developer's tool but things have changed a lot.
Also, Apple just announced their preview for new Siri AI capabilities and part of that is they seemingly are going to add MCP to apps (if developers of the apps want to) as tools, which would enable Siri to use functions of apps, similarly to Claude Desktop using functions of services through MCP tool calls.
I'd be tempted to try something similar for this season. But this is what I generally pay HR Block for; they should be relatively vetted (granted i don't know this for a fact).
like do you have tax accounting experience or do you intend to have someone who does help validate test returns to make sure they look right?
reply