> I could only get about 80% of what I wanted together with my hacky workflow.
I literally did this yesterday with solid results using Codex CLI. I used xhigh thinking and gpt 5.5.
I had it use KiCad directly via cli rather than via MCP, and I did make Claude Opus review it's work after every round. I got what I think will be a working revision A in about 10 hours of tinkering spread over a few days.
I'd honestly love to see the PCB. Using an LLM for a mostly geometric task like PCB layout feels like using a hammer to cook a chicken, unless KiCAD has some kind of text-based description language i'm not aware of that gets around having to specify coordinates.
KiCad schematic and board files are all text based with a fairly strict grammar, so you would be able to feed it directly into an LLM. Not that that means the LLM could actually make sense of it. Never tried though XD
This is the part everyone seems to forget. Any "new" jobs would be shitty low paying jobs, and it would mostly instead need to be automation.
Tariffs transfer wealth to the 1% and leave shit jobs that pollute the environment, which also happen to raise the cost of all goods, for everyone else.
The trope about external consultants is that your VP brings them in to review the company, and they talk to everybody and write a report on how to improve the business, and the report says exactly what you've been telling your VP but they've been ignoring you.
They are paid to justify decisions executives have already made. It's often referred to as due diligence, but in practice these reports mostly just allow executives to tell the board it wasn't their fault if it goes wrong.
To me, English is just another programming language. Some of us will always be better at it than others, and the ones that know other programming languages well will always have an advantage over those who do not.
When you are good at it there can be craft in it still.
A programming language is a formal intermediate language for turning human comprehensible instructions into machine instructions by means of an interpreter or compiler. We've now allowed that intermediate language to be English, because that's preferable to most people, and the "compiler" has become very complicated indeed as a result of that.
You still have to be able to express what it is you want in a way the machine can understand, it's just both simpler and less deterministic now.
This. Just because an llm can translate any language into a programming language doesn’t suddenly make all languages programming languages. Until I can ‘brew install englishc’ and so on, it’s not a f**ing programming language.
Can you define programing language in a way that includes all the current programming languages and excludes English? I kind of doubt it unless you just define it as "anything that isn't a human language", which would be silly.
Natural language is full of ambiguities and redundancies which makes it a poor fit for a programming language, which is why it is never used as such.
You don’t need a precise definition of a term to know what a thing is and isn’t (Wittgenstein has taught us that much at least). We just need to know that programming languages are used to express an executable computer programs (usually by translating to simple machine instructions) and that a natural language has never been used in this way in a significant manner.
A case in point. I bet you can‘t find a definition for a fish which includes cods and sting-rays, but excludes dolphins and shrimp. And similarly the IAU were unable to come up with a definition of a planet which included Pluto and Mercury but excluded Ceres and Sedna.
> Natural language is full of ambiguities and redundancies which makes it a poor fit for a programming language, which is why it is never used as such.
I mean, a quarter century ago Dijkstra argued your point compellingly, and he was right back then. If you read his "On the foolishness of “natural language programming”" (1978) you'll find that all of his most compelling arguments are gone now. Things have changed, and the machines can now largely cope with the ambiguity of language as well as the average human being can.
Since human language is the original source for the specifications we turn into formal code most of the time anyhow, we're really just asking if that original specification the programmers turn into formal symbolism is a form of code or not, and whether a good spec is equivalent to good code. I think it's difficult to argue that it's not, especially given that we now have these handy Natural Language to Formal Symbolism compilers.
> We just need to know that programming languages are used to express an executable computer programs (usually by translating to simple machine instructions) and that a natural language has never been used in this way in a significant manner.
I did that like 30 times today. Maybe it wasn't in the past, today it is. The path is now Specifications->LLM->Formal Symbolism->Machine Code, it used to be Specifications->Human->Formal Symbolism->Machine Code. The inputs and outputs are the same, and I would argue that the process is still "programming" regardless of syntactic games with semantics.
Eventually we'll find a more efficient version of that formal symbolism and stop using code designed to be human readable at all. Still nothing will really changed besides the input method.
You did no such thing. You fed some text into a statistical machinery which was able to infer another text from it. The first text just so happens to be a natural language and the inferred text was a formalized programming language which the statistical model had had its weight tuned to produce.
Statistical inference is a completely different process then compilation. Inferring is a completely different verb from compiling. Two different verbs which mean different things.
If we take your logic and explore its implication, we can just as easily claim that a project manager writing JIRA ticket is programming, and that JIRA is a programming language. The project manager wrote a ticket in natural language which was picket up by a developer who translated it (by your defintion of translation) to a formal language which got compiled to machine instruction and executed by a computer. This is obviously silly. And as silly as you find my description, I find yours equally silly.
Security industry going to be okay - someone will always pay for 0-days. If vendors wont pay its just gonna be US agencies, Israel resellers, China or Russia.
If you don't feed your army, you will soon feed someone's else's.
These days corporate security treats these workstations like a dummy terminal. No secrets live on the workstation. You have to re-auth with sso constantly with biometrics and are basically editing data that is in a cloud. So the risk to a corp is minimal where even in the worst case they are insured.
Zero days like this are being disclosed regularly so the idea of securing a windows workstation is tantalizing but you'll never feel satiated trying to drink that water so don't even try.
So yea there's plenty of windows users but we're certainly not hosting anything important on those boxes and would frankly be aghast at the suggestion.
> These days corporate security treats these workstations like a dummy terminal
Correct, "zero trust" is the buzzword but this is how Microsoft even recommends you set up your endpoint infra. Assume breach, treat every endpoint as if it is currently compromised or could be at any time. Laptops are basically ephemeral, when set up right, and can be wiped and re-imaged within an hour or less.
That's not unique to Windows either, that's how all employee/user endpoints should be managed.
Goodharts law. The metrics were always measuring the wrong thing, and now that we've finally optimized for the wrong thing successfully management will be forced to admit it and move on to another, slightly different, metric that doesn't actually equate to shareholder value.
It doesn't matter what the line actually measures, just that it goes up.
I found this claim interesting so I looked into it. Everything I can find shows that the intuition is accurate.
Companies with EOSP programs outperform those that do not in the market by about 17%.[0] Companies that perform layoffs, despite short-term stock boosts, underperform on a period of years showing a 14% decline in their Return on Assets (ROA) in the years following the layoffs.[1]
Synthid is a watermark which indicates the video is AI-generated, not a digital signature indicating it's real. Completely different use case and threat model.
I'm not aware of any secure digital signature schemes that don't require the thing they signed to be bit-for-bit identical to pass verification. There are perceptual hashing algorithms that could theoretically be used to build such a scheme, but such hashes are not second preimage resistant, so someone could create a modified video that still passes signature verification.
I'm not even sure what you're hypothetically describing here. You want a system that authenticates that an image hasn't been manipulated, but which still allows you to compress and transcode it? It's feasible in the same way that synthid was feasible, but I don't think there's actually a use case for it. You either want it to be unedited or you don't. I'm not sure how you can say "edited, but only exactly this much editing is allowed."
I suppose the validator could do a fuzzy match and just output a similarity score that compares the result to the original image. IE - This image is 75% similar to the original with something like a perceptual hash. Then it's the users problem to decide if 75% is close enough for their trust.
You want it visually identical but not necessarily bit-for-bit identical. Compression and transcoding should not cause validation to fail unless the compression artifacts are particularly severe, but even a tiny, one-pixel change that substantially alters the appearance of the photo should cause validation to fail.
And yes I agree this is hard to quantify and impossible with existing algorithms, that was my point.
I think some of them are actually run by Reddit directly. They couldn't find any way to keep making 'line go up', so they decided they could sumulate growth by machine translating Indian users to English and vice versa.
I think they're translating between users transparently to make it look like it's not a ghost town, and the machine translation reads like bot text.
Yes. RFK Jr. was corruptly placed in charge of America's healthcare by Trump in exchange for dropping out as a presidential candidate.
He is a conspiracist with no medical credeentials, and he believes, without evidence, that seed oils are response for most of the ills of mankind, Tylenol causes autism, SSRI's should not be prescribed, etc. None of his beliefs are mainstream or evidence backed, but he now has a huge megaphone.
I literally did this yesterday with solid results using Codex CLI. I used xhigh thinking and gpt 5.5.
I had it use KiCad directly via cli rather than via MCP, and I did make Claude Opus review it's work after every round. I got what I think will be a working revision A in about 10 hours of tinkering spread over a few days.
reply