HNHacker News
TopNewBestAskShowJobs

ivan_gammel

5,676 karma · joined February 28, 2011

Scaler CTO, Berlin, DE
submissionscomments
ivan_gammel··on Something is afoot in the land of Qwen
Yes. 996 is for lazy people.
ivan_gammel··on Something is afoot in the land of Qwen
Having one „champion“ is flawed European approach. We need local competition and headhunting to make it fly.
ivan_gammel··on AfricaMuseum refuses to yield Congo geological archives despite US pressure
If understand this correctly, the museum is currently digitalizing the archive, which is then expected to be publicly available. United Oligarchs of America want to get the papers for „digitalization“ in their own hands, which will be de facto exclusive access to the information contained in them. It smells like neocolonialism and corruption.
ivan_gammel··on Something is afoot in the land of Qwen
Offering „You are welcome“ relocation package to Anthropic might be a good idea.
ivan_gammel··on Claude is an Electron App because we've lost native
It might be a strange thing to say, but Java is still viable alternative route. You can build a nice and fast cross-platform desktop application on it today. The language was designed for this kind of things. The entry barrier is quite high though.
ivan_gammel··on Switch to Claude without starting over
> We should just have language / technology specific models with a focus on recent / modern usage.

The “just” part is a big assumption. It is far from easy, given that modern best practices are always underspecified. The effective model for coding must have reasoning signals to be much stronger than coding patterns, and that, I suspect, requires very different architecture.

ivan_gammel··on Switch to Claude without starting over
Greenfield implementation is not flawless as well.
ivan_gammel··on Switch to Claude without starting over
The more code is out there, the worse is the average in the training dataset. There will be legacy approaches and APIs, poor design choices, popular use cases irrelevant for your context etc that increase the chances of output not matching your expectations. In Java world this is exactly how it works. I need 3-5 iterations with Claude to get things done the way I expect, sometimes jumping straight to manual refactoring and then returning the result to Claude for review and learning. My CLAUDE.md (multiple of them) are growing big with all patterns and anti-patterns identified this way. To overcome this problem model needs specialized training, that I don‘t think the industry knows how to approach (it has to beat the effort put in the education system for humans).
ivan_gammel··on The whole thing was a scam
This interpretation is kinda obvious to anyone who has seen similar schemes in other countries. It‘s done almost by the book, except there‘s no criminal case against Anthropic management or shareholders, because USA is not yet there.
ivan_gammel··on OpenAI agrees with Dept. of War to deploy models in their classified network
Another plausible explanation that is familiar to a lot of people in other countries is banal corruption. Kick out one competitor on bogus allegations, then on the next day invite another one… what else that could be?
ivan_gammel··on Tell HN: MitID, Denmark's digital ID, was down
Does any government in the world issue only digital IDs?

There‘s always possibility to have your travel passport as a backup (and when traveling abroad your domestic ID is suitable for recovering passport).

ivan_gammel··on Tell HN: MitID, Denmark's digital ID, was down
Temporarily losing access is just inconvenience. Imagine the same but you lost the wallet with your only cash and your passport in pre-digital times, you are far from the nearest embassy and nobody understands your language. You are fully at the mercy of the locals and your money aren‘t coming back.
ivan_gammel··on IDF killed Gaza aid workers at point blank range in 2025 massacre: Report
That‘s just two flavors of nationalism, an outdated idea from last couple of centuries that leads nowhere. The only reasonable way is federalism and deep political integration, where Israel, Palestine, Lebanon, Syria, maybe Jordan build a common state with free trade, free movement etc. That may sound a distant future at best, but the alternative is the same bloody status quo.
ivan_gammel··on IDF killed Gaza aid workers at point blank range in 2025 massacre: Report
Nah, let’s not let them to set the narrative. It is not antisemitism to criticize Israel and I do not care what Israeli or my (German) government says about it.
ivan_gammel··on IDF killed Gaza aid workers at point blank range in 2025 massacre: Report
Whether you agree with it or not, does not matter. It is ambiguous due to a simple fact that I did not had the choice of interpretation in my mind. It is how I understood it and it differs from your understanding. The author should have been more clear.
ivan_gammel··on IDF killed Gaza aid workers at point blank range in 2025 massacre: Report
The message is ambiguous. It can be interpreted the way I read it.
ivan_gammel··on IDF killed Gaza aid workers at point blank range in 2025 massacre: Report
it‘s „some“, not „all“. Religious extremism by definition.
ivan_gammel··on IDF killed Gaza aid workers at point blank range in 2025 massacre: Report
You are surprisingly right. I know people who served in IDF and would prefer to have nothing in common with those criminals. Generalizing to them would be wrong. It is not voluntary service, different people are required to serve. But people aside, is IDF as institution rotten? It is not generalization to say „yes“, when such things happen. An institution is an entity with the agency to prevent such things and not only did it fail, it covered up. Is Israeli government complicit? Hell, yes, same reason.
ivan_gammel··on IDF killed Gaza aid workers at point blank range in 2025 massacre: Report
Then I apologize without retraction.
ivan_gammel··on IDF killed Gaza aid workers at point blank range in 2025 massacre: Report
No. Your message is. A lot of people commit mortal sin of logical fallacy by extending the responsibility for actions of certain group of people to everyone sharing with them ethnicity or religion. It‘s the stupidity worth of the strongest condemnation given the context.

It‘s not jews committing war crimes in Gaza, it‘s zionists. It‘s not muslims or Palestinians planning and executing terrorist attacks, it‘s religious extremists and far right nationalists. When there will be common understanding of this simple truth, fighting the root causes will be much easier.

ivan_gammel··on How Taalas “prints” LLM onto a chip?
> users report that "model got dumber"

I have seen this with ChatGPT progression from 4o to 5.2 applied to the newest model. Old prompts stop working reliably, different hallucination modes etc.

ivan_gammel··on The Age Verification Trap: Verifying age undermines everyone's data protection
If third party does verification with ZKP, you only need to trust that third party. The company that requires verification will not have any data to store.
ivan_gammel··on How Taalas “prints” LLM onto a chip?
>For a 2.5 kW Server?

I don’t see any reason why this should not drop to 100-300W at peak with maybe 100W*h of daily usage on smartphones.

ivan_gammel··on How Taalas “prints” LLM onto a chip?
More expensive than what? How much equivalent low latency inference costs today?

I think you completely miss the UX point here. In 1997 CRT screens were mainstream, LCD was in the early stage, phones had antennas. In 2007 an iPhone with LCD touch screen changed the UX of computing forever. This tech that we see today is a precursor of technology that will dominate tomorrow. Today local inference is painful and expensive, it consumes a lot of energy. NPUs/GPUs solve nothing here, and they will always be less effective than hardwired models - by design. So only question is, when the consumer performance expectation for open-weight models will cross the price curve of specialized chips. It may happen earlier than for generic NPUs.

ivan_gammel··on How Taalas “prints” LLM onto a chip?
I feel weird defending Taalas here, but this argument is quite strange: of course it is more expensive now. It is irrelevant - all innovations are expensive at early stage. The question is, what this technology will cost tomorrow? Can it do for consumers what NPUs could not, offering good UX and quality of inference for reasonable price?
ivan_gammel··on How Taalas “prints” LLM onto a chip?
Most consumers do not care about reprogramming or fine-tuning and have no idea what NPU is. For many (including specifically those who still mourn dead AI companions, killed by 4o switch) the long term stability is much more important than benchmark performance of evergreen frontier model. If Taalas can produce a good hardwired model at scale at consumer market price point, a lot of people will just drop their AI subscriptions.
ivan_gammel··on How Taalas “prints” LLM onto a chip?
I don’t see why. Today frontier models are already 2 generations ahead of good enough. For many users they did not offer substantial improvement, sometimes things got even worse. What is going to happen within 1 year that will make users desire something beyond already working solution? LLMs are reaching maturity faster than smartphones, which now are good enough to stay on the same model for at least 5-6 years.
ivan_gammel··on How Taalas “prints” LLM onto a chip?
Why are you thinking about phones specifically? Most heavy users are on laptops and workstations. On smartphones there might be a few more innovations necessary (low latency AI computing on the edge?)
ivan_gammel··on How Taalas “prints” LLM onto a chip?
The generation of frontier models from H1 2025 is the good enough benchmark.
ivan_gammel··on How Taalas “prints” LLM onto a chip?
When output is good enough, other considerations become more important. Most people on this planet cannot afford even an AI subscription, and cost of tokens is prohibitive to many low margin businesses. Privacy and personalization matter too, data sovereignty is a hot topic. Besides, we already see how focus has shifted to orchestration, which can be done on CPU and is cheap - software optimizations may compensate hardware deficiencies, so it’s not going to be frozen. I think the market for local hardware inference is bigger than for clouds, and it’s going to repeat Android vs iOS story.
← PreviousPage 7 of 34Next →