Not hard, but time consuming. In the past two weeks I've had Claude Code write me around 35k lines of code across 350 commits. It's a project which is giving positive impact to the company, but we would never have started it without CC as the effort would have been too big compared to the impact.
I'm an avid user of the Claude Code planning feature and I like how Claude Code asks for questions. I also often iterate the plan before finally giving it a go.
How do you solve this in Kelos?
I tried to check the code base, but it didn't really provide any glues. I guess I could instruct the agent to build a plan and to post the plan in the issue and then iterate that with written comments in the issue. Is that how you run it?
> I guess I could instruct the agent to build a plan and to post the plan in the issue and then iterate that with written comments in the issue. Is that how you run it?
Exactly.
I simply just create an issue if there's a simple bug and let Kelos handle it.
If there's a complex issue that I'd like to build a plan for before implementing it, I build the plan locally and let the agent create a github issue for it, then Kelos will handle it.
But there are companies which are only serving open weight models via APIs (ie. they are not doing any training), so they must be profitable? here's one list of providers from OpenRouter serving LLama 3.3 70B: https://openrouter.ai/meta-llama/llama-3.3-70b-instruct/prov...
Not sure about that. My Android warns me about my wife's airtags so often, that if I would actually be tracked by a malicious airtag, I would just assume it's one of my wife's tags. This could be prevented if I could mark a tag to be trusted on my Android phone, but no such feature exists.
I was just helping my dad with a brand new Lenovo laptop with Windows 11. It felt unbelievable slow and sluggish. Just opening file manager to create a new folder lagged so much it felt like this would have been a 15 years old computer.
While I personally use Ubuntu on my laptop for several years now, when I helped my relative with a brand new laptop (huawei) with Windows 11 I was suprised how fast it was despite being very cheap, I don't remember any version of Windows that had such a performance, at least visually.
Out of curiosity, what model does your father have?
To me it's generally pretty quick outside of File Explorer. The reskinned File Explorer is an absolute car crash - it's like they've taken everything that made 2005 KDE awkward and bolted it on top of Windows.
Hopefully you can write the teased next article about how Feedforward and Output layers work. The article was super helpful for me to get better understanding on how LLM GPTs work!
I remember having this argument with my professor at the school, who insisted that a function should have only one "return" clause at the very end. Even as I tried, I could not get him to explain why this would be valuable and how does this produce better code, so I'm interested on hearing your take on this?
It helps prevent bugs with state. The apple login bypass bug comes to mind.
Basically, you have code in an "if" statement, and if you return early in that if statement, you might have code that you needed to run, but didnt.
Forcing devs to only "return once" encourages the dev to think through any stateful code that may be left in an intermediate state.
In practice, at my shop, we permit early returns for trivial things at the top of a function, otherwise only one return at the bottom. That seems to be the best of both worlds for this particular rule.
> In practice, at my shop, we permit early returns for trivial things
Are you also writing C or similar? If so, then this rule is relevant.
In modern languages, there are language constructs to aid the cleanup on exit, such as using(resource) {} or try {} finally {} It really does depend on if these conveniences are available or not.
For the rest of us, the opposite of "no early return" is to choose early return only sometime - in cases where results in better code, e.g. shorter, less indented and unlikely to cause issues due to failure to cleanup on exit. And avoid it where it might be problematic. In other words, to taste.
> Kent Beck, Martin Fowler, and co-authors have argued in their refactoring books that nested conditionals may be harder to understand than a certain type of flatter structure using multiple exits predicated by guard clauses. Their 2009 book flatly states that "one exit point is really not a useful rule. Clarity is the key principle: If the method is clearer with one exit point, use one exit point; otherwise don’t".
There are missiles in which the allocation rate is calculated per second and then the hardware just has enough memory for the entire duration of the missile's flight plus a bit more. Garbage collection is then done by exploding the missile on the target ;)
reply