I really wish seemingly intelligent people would stop using the abstraction analogy (like the article does). The key word is: determinism. Every level of abstraction (inc. power tools, C, etc.) added a deterministic layer you can rely on to more effectively do whatever it is that you're doing - same result, every time. LLM's use natural language to describe programming and the result is varied at the very best (hence agents, so we can brute force the result instead). I think the real moat is becoming the person who can actually still program.
People always say this but it’s misguided imo. Yes LLMs are not deterministic, but that’s totally irrelevant. You aren’t executing the LLMs output directly, you’re using the LLM to produce an artefact once that is then executed deterministically. A spec gets turned into code once. Editing the spec can cause the code to be updated but it’s not recreating the whole program each time, so why does determinism matter?
In my experience, I'm using LLMs as my abstraction to "junior engineer". A junior engineer isn't deterministic either. I find that if you treat the LLM output like a person's output, you're good. Or at least in my projects, it's been very successful. I don't have it generate more code than I can review, or if I give it a snippet to help me fix it, if it ends up re-writing it like an ambitious engineer would do, I tell it to start over and make minimal changes.
I guess I'm not spun up about the determinism because I've been working at the "treat it like a person" level more than the "treat it like a compiler" level.
To me, it's really like an engineer who knows the docs and had a good memory rather than infallable code generator.
I work at a small company, so we don't have tons of processes in place, but I imagine that if you already had huge "standards" docs that engineers need to follow, then giving the LLM those standards would make things even better.
The thing is you can quickly teach a Junior how to respect a specification contract, so that with very minimal oversight, you get the wanted implementation. And after a few years (or months), the communication overhead get shorter. What would have been multiple rounds of meetings and review sessions are a short email and one or two demos.
What I've been learning as a 20% "harness engineer" is that in order to get the models to "learn" you need to add both documentation and static checks, as well as often custom skills. My main project at work has issues where the AI will often get super confused and step on itself trying to run tests - so the answer is writing better docs (AGENTS.md) and providing deterministic tools to work with the projects.
Large software projects (I'm thinking google3) often have large amounts of both of those things, as they're always getting new developers joining.
> Reviewing code is still way faster than writing code.
Writing code results in a much better understanding of the code than reviewing it
In fact I would say that in large complex codebases, in order to develop the same understanding of what the code is doing might actually take longer than writing it from scratch would have
this is the way LLMs _should_ be used, as an assistant to create reliable, deterministic code. and honestly, they're fantastic when used this way. build the thing you need with the LLM, then put the LLM away.
but in practice, the current obsession with agents means people are creating applications that depend entirely on sending requests to LLMs for their core functionality. which means abandoning the whole idea of deterministic software in favor of just praying that all of the prompts you put around those API requests will lead to the right result.
try distributing this spec amongst your team members, ask each of them to drive it to completion. no follow up edits. deploy to individual environments and then run a rigorous test suite against all of the deployments. see if all of them behave the same way.
I grant that there's a definition of abstraction that LLMs don't fall into. But people describing LLMs as another abstraction layer aren't all misunderstanding this. Instead, they are using the term ... more abstractly.
EG: How did Mark Zuckerberg make software five years ago?
He's as capable of opening up an editor as I am, but circumstance had offered him a different interface in terms of human resources. Instead of the editor, he interacts with those humans, who produced the software. This layer between him and the built systems is an abstraction, deterministic or not.
Today, you and I have a broader delegation mandate over many tasks than we did a few years ago.
The way I frame this is that LLMs are not replacing the tools, whic are are deterministic. They are replacing the humans, which are themselves non-deterministic, as in your Zuckerberg example.
LLM's don't have to achieve perfect reliability to replace lots of work. They just have to reach the balance of reliability and cost suitable for a given task. This will depend on the task.
I see what you're getting at, but determinism isn't the right word either. LLMs are fundamentally deterministic -- they are pure functions which output text as a function of the input text and the network parameters[1]. Depending on your views on free will, it could be effectively argued that humans are deterministic as well.
The concept you're touching on is the idea that LLMs (and humans) are functions which are inscrutable. Their behavior cannot be distilled into a series of logical steps that you can fit in your head, there are no invariants which neatly decompose their complexity into a few interpretable states, and the input and output spaces are unstructured, ambiguous, underspecified, and essentially infinite. This makes them just about impossible to reason about or compose using the same strategies and analysis we apply to traditional programs.
[1] Optionally, they can take in a source of entropy to add nondeterminism, but this is not essential. If LLM providers all fixed their prng seeds to a static value, hardly anyone would notice. I can't imagine there are many workflows which feed an LLM the exact same prompt multiple times and rely on the output having some statistical distribution. In fact, even if you wanted this you may just end up getting a cached response.
Let's be real, if you and I both ask claude to generate a feature on the same project, what are the chances that it spits out 100% replicated code? But if we are to build the project using a Dockerfile, we will get the same binary and the same image. Products around LLMs are non deterministic unlike compilers.
I can assure you that a fully deterministic and equally effective claude is possible to build. And yes, that would mean identical prompts would yield 100% identical output 100% of the time. It would still make the occasional logical or factual error, but it would do so deterministically. Would this solve any of the problems with building reliable programs using LLMs?
it's nondeterministic because we chosen it by having higher 'temperature' in settings. I bet if you run open weights model with temperature 0 and on the same device the same prompt and turn off parallelism you will have more deterministic result (excluding some floating point operations).
> Optionally, they can take in a source of entropy to add nondeterminism, but this is not essential. If LLM providers all fixed their prng seeds to a static value, hardly anyone would notice
Everyone added /dev/random to their offerings, so every LLM tools for coding are non deterministic.
There's something to be said about the fact that the very people who would use deterministic layers to build stuff are... non-deterministic. We, as humans, have our set of pros and cons, wins and failures. Even the most brilliant coders on earth will make mistakes from time to time. I often fail to see this getting accounted in any conversation when there is a critique towards LLMs, as if we humans are not flawed in our own ways, with a huge degree of variance across individuals. Good and bad code existed prior to LLMs. If you're hiring someone to do code, you're basically using some heuristics to trust this person will do a good job. But nothing is ever guaranteed 100% deterministically ever. Without thinking it that much, LLMs will sometimes produce better code and manage systems that some people who are earning salaries out there. Possibly sub-par developers if we were precise, but professionals in the meaning of the word (that are being paid to do work).
At the end of the day, what matters is how willing the person behind a given task is when it comes to deliver quality work, how transparent and honest they are, to understand requirements, and a pleasure to work with along other humans. AI/LLMs are just extra tools for them. As crazy as it might sound, but not so many people are willing to push boundaries and deliver great work. That is what makes the difference.
Are other layers truly deterministic though? Do I know for sure whether that object has been garbage collected or not? How many cycles this instruction will take to run?
Spotify DJ is terrible - just like their Apple TV app (which is perpetually buggy). However, their radio mix feature is so good that it's become the #1 reason I can't switch.
A Figma competitor is a really good thing - many people are waiting to jump ship for a decent competitor. However, for the sake of longevity & lawsuits, maybe tweak the UI a little bit. Just to make it not look exactly like Figma. Well done though!
I wish Europe would just push back. More than what they are currently. There is so much potential there, but somehow the EU all look at the US as some form of idealogical father figure. Excuse the hyperbolic-talk.
I don't think it is true. It's like saying "I wish those kids didn't let the bigger one bully them". The reason the bully is bullying is because he is in a position to do it.
The EU is being careful because the US are more powerful.
Trump has repeatedly backed off when he's challenged. It's happened time and time again. It's the reason TACO is a thing. The best strategy against him is to be relentless about pushing back, even if on paper the US is more powerful.
It seems you can also just lie to help him save face, like Canada did when it agreed it would adopt very strict border control policies to stop "drugs coming into the USA," and listed out steps that all were just existent Canadian laws and policies.
The problem that US generals have right now is that Trump has gotten the idea that the US (viz., he himself, in his mind) ought to literally own Greenland and he does know how real estate works. Treaties, mineral deals, guarantees for additional military bases that would mean de facto control over Greenland would work with a rational person. However, they won't work with someone who insists on buying or annexing a country to own its territory.
Yeah, another strategy is to just give him something he can claim as a W even if it's bullshit, or to glaze him enough. He's so hyperfixated on owning Greenland though, that I'm not sure those will work this time.
This effectively means the end of the 0 percent tariff on US products. There are also already calls in the EP to activate the Anti-Coercion Instrument:
The problem is NATO, a lot of the EU is reluctant to push back because at the end of the day the US guarantees that Russia cannot pursue the type of landgrab it is currently trying to do in Ukraine against other states. The risk that the US runs into when trying to take Greenland is that this argument loses weight instantly, so the expectation is that the EU will be much more willing to use its anti coercion tools if Trump tries to make it a reality.
Russia already fails in Ukraine where they are fighting with our old junk, and the other EU States are kicking their defense industry in full gear. What makes you think they could win a full scale war against the EU
Russia don't have to be able to win a full-scale war against the EU for such a war to break out, it suffices that deterrence breaks down sufficiently that Russia get the idea they can get away with some land grab, e.g. in one of the Baltic countries.
The war in Ukraine illustrates very well the difference between perception and reality. Perception counts for deterrence.
The Baltics are protected under the EU defence clause, NATO or not they will be assisted by the EU.
It's already quite clear the US has virtually left NATO, at this point they wouldn't assist at all with a landgrab in the Baltics so I'm glad the EU defence treaty is more forceful about the level of aid/assistance than Article 5.
NATO at this point is virtually dead, there's no trust in the USA and the rhetoric about Greenland has cemented it. Hope the Canucks can join a defence pact with the EU, the Trump admin and its Project 2025 achieved what they wanted.
You might want to check some reliable sources about how the war is going for Ukraine, because it seems like you think they are kicking some Russian bootie, which is simply not the case. Take the US (and risk of mutual worldwide nuclear contamination) out of the equation and Ukraine would be in even worse shape.
> because it seems like you think they are kicking some Russian bootie,
This seems like quite the assumption.
It is generally a mistake to attempt telepathy/IP.
Russia not doing nearly as well as one might expect for an aging out core of a former superpower is not an equivilance with their target is kicking their arse.
The grind Russia is having to go through against Ukraine is an indicator of how it might fare against a full NATO (sans the US).
".. because at the end of the day the US guarantees that Russia cannot pursue the type of landgrab it is currently trying to do in Ukraine against other states"
I am sorry to say that we (Europeans) increasingly do not believe that the US would help us.
It's like when every liberal scoffs at leftists opposing US imperialism, nothing about the power balance has changed. Europe was always a vassal of the empire. This is the liberal international order, this is what that means, not what they tell you it means, but what it actually means.
That's why they can kidnap Maduro, have the BBC censor the word "kidnapped" in their reporting on it. Have every European politician applaud it, point to Maduros case against him at the ICC and have Netanyahu fly over France. You can't do anything about Greenland, the same way you can't do anything when he comes for Norways state-owned extraction industry next. Liberals can scream hypocrisy tears all they want, this is the world they built. The empire is coming home.
Leftists wanted her to not be a dog shit politician in order for her to win, they were screaming for her to embrace real substantive policy positions and not business as usual, corrupt, liberal elitism. The same leftists are now in the street protecting communities from the gastapo, while liberals debate about which words they can say. It were those exact liberal politics that lost Hillary the election too, and then you were screaming too about how it was all Bernie's fault. For christ sake, Trump was able to sell himself as the PEACE CANDIDATE, how can you fuck this up so badly?
Because when you have a brain you understand that a more center oriented candidate with Luke warm opinions in policies has more chances of being acceptable to a larger audience than a candidate with more "substantive" policies.
Having Biden running at the start was the real issue.
I just don't understand the perspective that Trump is a historical threat and therefore we can't accept business as usual. I have a number of disagreements with the status quo myself, but I'm not going to pursue them until Trump is out of power, because I want to absolutely minimize the number of people who feel they have to choose between supporting Trump and abandoning some principle of theirs. To me, any other strategy seems tantamount to saying that Trump isn't so bad.
But your lukewarm candidates lost twice, Hillary lost, Kamala lost. The point we are making is that they lost, because they are lukewarm. There is a reason Trump won in the first place you are ignoring, you are ignoring the times of unprecedented grievances that people have, people want real change. Trump represents that change to people, a fascist lie and scapegoating of course, but you are representing the comfortable elite under whom nothing will ever change for the better for anyone. All you have is complain about leftists, we didn't loose, you lost twice. Dems are more unpopular than ever, even now under fucking Trump, your politics are dogshit and you don't have anybody else to blame for it.
I don't represent or subscribe to what you think! I agree that both of them were weak candidates who lost where a better candidate could have won, and I myself have been growing away from the Democratic party ever since the 2016 primary.
What I cannot agree with, what I find completely unacceptable, is the idea that any dispute over candidate quality can justify splintering the anti-Trump coalition. If Bernie were the 2024 candidate, I assure you I would have even harsher words for any business types who ran around complaining about him.
Two party system is such a mess. I blame two party system more than anything. When you reduce everything to two party, its so reductive and this is the mess you are gonna have to face because of it.
A key point is that it's an electoral system from hundreds of years past that was never intended to be a two party system, one set up by founders who in the large wre not even fans of party politics (one, two, or more).
It is a system that by it's design is more or less doomed to iterate into a Hotelling's law quagmire of two nose to nose opposing sides neither of which represents any kind of majority or popular view.
The US electoral system is well past due for a revamp, as recommended by it's founders who judged it "good enough" for a while ... until a despot appeared.
I agree but trust me when I say this, its not gonna happen.
I see people so entrenched in American politics who cant believe that there can be independents atleast in how the current voting works
They probably need better voting mechanisms... but for which they are gonna have to vote and no republican or democrat is gonna propose this ( i really don't think so) and the people can probably only vote for republican or democrat (independents very few) in the current system...
So its doomed and this is the reason why. A lot of American politics in the end feels like this or that, not knowing the nuances and polarization (in some sense) from both parties while still bieng the same (corruption stemming from lobbyists)
It's just really sad to see because to me its like not just Trump being a hostile takeover (which he is) but rather that both parties and the system failed the people so that someone like trump could spin up in the first place and now this is even happening.
If I were to tell you even 2 years ago all the things happening in America, you would believe we are in a black mirror episode or Its a bad dream but its reality now & we (non Americans) just gotta deal with America impeaching on other countries sovereignity trying to buy things outright and all escalations and the final one remaining is war and they haven't put it off the table as well
As a non American you have a semi reasonable chance of being to sit back, take a beat, and watch (maybe) Trump implode and self destruct within the US system and maybe some following rebuilding of the system "as intended" with better safeguards.
Trump’s triumphant narrative is not working at home, either. A new CNN poll released Friday shows that fifty-eight percent of Americans believe that Trump’s first year in office has been a failure. Americans worry most about the economy, but concerns about democracy come in second. The numbers beyond that continue to be bad for Trump. Sixty-six percent of Americans think Trump doesn’t care about people like them. Fifty-three percent think he doesn’t have the stamina and sharpness to serve effectively as president.
Sixty-five percent of Americans say Trump is not someone they are proud to have as president.
One of the issues I have with this is that a Weaker America does mean a better China and the reasons I have been vocal is that this doesn't have to do much with America itself but rather the fact that we need a multi polarized world in first place with (I think) policies of non alighnment because Europe aligned itself to US for the most part, its coming so much as a shock-wave in the first place.
A stronger China does mean more micro-agressions towards Asean countries in general (japan,south korea,India) and QUAD members (minus the united states) so it would be beneficial if the EU block could align itself with the members of Asean who still align with democratic ideals and similar.
This is probably why most countries officials (or people interested in geopolitics) are on the edge of their seats
A better China, or stronger China, would likely look like China as a trading powerhouse dominating Atlas of Economic Complexity rankings globally, exerting greater control over the bottleneck of China Sea through which almost all its inbound and outbound goods traverse.
It's unlikely (but possible) to see it flex as a global military powerhouse in the same manner as other great empires have done in the past - but it is probable that china will continue to extract "water resources" as food from Africa and elsewhere as it, the Saudi's, and others already do .. in China's case with the backing of its own mercanaries and with US mercanaries (they were hiring Erik Prince and Blackwater not so long ago).
This is a pattern the world has seen before - great powers come and go, meridans and global financial centres have moved before and will move again.
Yes, there has been an uneasy peace of sorts for 75 years or so, do be aware of and prepared for transitions.
Not sure if its the number of parties that is the primary issue. Corporate lobbying, campaign finance, bribery, and cultural distractions (intended to posit groups against each other) are some areas that concern me more.
The EU has a huge strategic problem because they let their own defenses and industry rot for decades and can't functionally stand alone against Russia, US pressure, and Chinese economic infiltration / industrial replacement at the same time. At least, not without great sacrifices the population isn't willing to make, like pension reform.
So they are playing gentle with the US because it's the least bad choice right now.
The EU is 450 million people! It's the size of the entire continent of south america! It was the richest part of the world for centuries! They absolutely should be able to function as an independent block with international trade for convenience and not survival.
Not even the US can stand against China by itself...
The EU still has a large military industrial base getting revitalised as we speak, it didn't rot, it simply didn't need to pump out massive amounts of gear until this point.
Poland, Germany, France, Spain, Italy, Sweden, Finland, Czechia, etc. all have different kinds of weapons manufacturers. You can even include the UK, and Norway in the mix even though they aren't in the EU.
No, the EU obviously did need to pump out massive amounts of gear, and failed to, and that's why four years into a war, Ukraine is still suffering under the yoke of a country with 1/10th the GDP of the EU.
If the EU had taken their responsibilities seriously given the MASSIVE THREAT next door, Ukraine would have had massive ordinance dumped on it in March 2022 and been free of Russians by Christmas.
It failed for political reasons. Political leaders being afraid to get involved in the war. Also do not rule out right wing political parties that are often anti-Ukraine and pro-Russia due to being sponsord by Russia.
The USA also has had it share of preventing the EU from getting involved.
China has a population of 1.4B people yet they import huge quantities of fuel and food and we can't pretend that they lacked investments in core industries.
That may actually be an advantage: position Europe as a neutral block that trades with everybody, and it may actually be valuable enough as a neutral that anytime one of those three has designs on it, the other two would naturally have to combine to thwart them.
The fact is that there is no potential there. Europe has no leverage over the US. It is not holding back anything, it has nothing.
Somehow when the US went to war with Russia, it ended up completing the conquest of Europe. Europe used to just be stagnant. Now it is stagnant and isolated from everywhere except the US, and the US treats it accordingly.
> ASML relies on the United States for several of its components, and it’s this very reliance that has allowed the United States to use the Foreign Direct Product Rule and impose export controls on ASML products. However, there are signs of a shift. ASML has already started to reduce its dependence on American technology, aligning with the EU’s goal of strategic autonomy. Earlier this month, ASML announced a major investment in Mistral, France’s flagship AI startup. The Dutch firm invested $1.5 billion in Mistral, becoming the company’s largest shareholder. The deal was widely seen by policymakers as a move that strengthens European ‘digital sovereignty.’ In a sector dominated by American tech giants, ASML’s Mistral investment represents a growing realization from Europe: cooperation within the bloc is necessary for the EU to stay competitive in the AI race.
---
I don't follow, how exactly does the investment into a French AI startup reduce ASML's "dependence on American technology"? Is it a supply-chain dependence, or a revenue-making dependence?
Recently I've put Claude/others to use in some agentic workflows with easy menial/repetitive tasks. I just don't understand how people are using these agents in production. The automation is absolutely great, but it requires an insane amount of hand-holding and cleanup.
I think a lot of modern day stoicism is stoicism-without-hardship. And I think hardship is necessary for stoicism - otherwise all you have is determined detachment, which is something else entirely.
Ton arrière-arrière grand-père a vécus la grosse misère
ton arrière grand-pere il ramassait des cennes noires
et pis ton grand-pere miracle est devenue millionaires
ton pere en na hériter il a toute mit dans ses réer
et pis toé tite jeunesse tu doit ton cul au ministaire.
pas moyen davoir. un prés dans une intitustion banquaire.
pour calmer tes envie de huldoper la cassière tu lit des livre qui parle ... de simplicité vonlontaire
- I think each generation can have a different reason for adopting any philosophy it’s about whether it serves you or not.
