Still can't seem to do RAG or otherwise answer questions about a specific URL. Seems kinda useless compared to even copilot or Gemini 2.5 unless I'm missing something.
I asked it to summarize this thread [0] and it just said "that thread is about the monthly "who's hiring"' on HN", not even close.
I don't really understand where this fits in the market? It's not as intelligent as the pack leaders and is about onpar with GPT-4o mini. When comparing it to GPT-4o mini further, while it is a bit faster, GPT-4o mini is a lot more cheaper [Source: https://artificialanalysis.ai/models/nova-pro/providers].
In terms of value for money, I would probably go with GPT-4o mini and not Nova Pro. Maybe Amazon feels that it needs to have their own offering to stay relevant?
My guess is it's also about enterprise agreements.
For many larger enterprises, governments, etc, the barrier to trying these things is to pay for them (new contracts, RFPs, etc).
But all of them already have enterprise agreements with MS, Amazon, etc. So there must be some class of customers to whom it's easy to just add this to their AWS bill.
Amazon Nova is a foundation model created by Amazon and is offered as one of the models you can use in AWS Bedrock, so the model gets a marketing page for it. Note that Llama also has an AWS marketing page (as do other models), but that doesn’t make them AWS products: https://aws.amazon.com/bedrock/llama/
Amazon Nova Chat is a different product that uses Nova, buts it not AWS. Notice that if you try to use Nova Chat, you log in using your Amazon.com account and not an AWS account.
These business can easily Anthropic models through AWS Bedrock. All it requires it a simple clickthrough EULA. That's what we do at the F500 non-tech company where I work. The same is true with OpenAI models in Azure.
I can't imagine AWS is going to get much usage of these models... but you have start somewhere I guess.
Of course a trademark of a common word might arise naturally from being in the market and becoming widely recognized (e.g. Apple), but you can't file a trademark for it until you have become widely recognized.
This is why Nova is a very popular term for products.
That's not right. The more a term is in the market, the more difficult it is to get a trademark approved.
The difficulty comes from the obligation to protect a trademark - if you have trademark rights to a term and don't take reasonable steps to protect it, you're at risk of losing your rights.
>> Amazon makes it easier for developers and tech enthusiasts to explore Amazon Nova
Well, you don't. Can't find a simple way to try it out. Not that i expected anything else. The only thing saving AWS in the AI wave seems to be their deal with Anthropic.
I have empathy for them. It's a complex product with many different use cases, so there is no one-best UI for anyone. Further, accommodating multiple skill levels is a genuinely hard problem.
That said, there is a larger problem: every team is incentivized to ship as fast as possible, and good UI takes time and consistency across every team. As a manager with a team that works closely with design systems, it's a brutally hard problem to get right and nobody wants to slow down in the design phase or have limitations placed on them for the purposes of consistency.
Eh, not really. They are behind in this 'race' on GPU stock as well as spare power, and I question if they will be able to truly catch up before the interest wanes.
GDPR is simply ahead of its time from an American perspective. Like Mobile Telephony in Europe when pagers were still a thing in the US.
In time the average american consumer will understand the monetary value of their PII and usage metadata and demand adequate protections - which effectively is all that GDPR does. Given the actions of the current cabinet, I feel we are in fact accelerating towards this inevitable outcome.
GDPR does more than demand protections. It demands data locality. It demands that encryption and access controls be done in a certain way. It backs up its demands (that are sometimes vague) with huge fines.
It really doesn't do anything more than mandate consumer side protection.
Data locality in legally compatible jurisdictions is the most fundamental form of protection there is. Without concepts such as Safe Harbour and data locality, handling of PII would be farcical amongst MNCs.
Re: Demands on Encryption? The most prominent mention of encryption is in Article 32(1)(a), which mentions the “pseudonymisation and encryption of personal data” as measures that organisations can adopt.
However, it is important to note that encryption is not compulsory. Instead, the GDPR takes a risk-based approach, meaning that the decision to encrypt data depends on the sensitivity of the data, the risks involved, and the potential impact on data subjects.
Backing up demands with fines is about the only way consumer protections are realised as corporate mandates rather than friendly advisory. Name me another comparable legislation that achieves its goals without resort to punitive measures for non-compliance?
In short, you would far better understand the intent, purpose, and reality of GDPR if you engaged with it as a piece of vital EU consumer protection legislation, rather than some sort of draconian shake-down of American Capitalist practices.
Ah excellent. Encryption is not compulsory, but doing a bureaucratic risk assessment of whether you need encryption is. That is so much less work.
In reality, GDPR is a jobs program for eurocrat auditing and consulting firms combined with an effort by Facebook and Google to prevent European competition. Note that GDPR fines are big enough that they can crush a small company, but small enough that Google wouldn't care.
Doing a risk assessment to determine TLS and encryption requirements is a fairly fundamental part of handling of any personal information in any context. That GDPR puts a basic framework and foundational expectations around this as a legal mandate can only be a positive thing from the POV of the consumer.
The notion that this is either a consulting gig fix or an effort to prevent European competition is naive and farcical in the extreme. The three highest fines for Meta (1.2b, 405m, 390m) total €2 Billion. More than every other GDPR fine combined.
Note that GDPR fines for individuals and SMEs are in the 3 to 5 figure range, and come under very basic grounds following repeated warnings. The intention is not to 'crush' anything, least of all SMEs in a globalised marketplace.
This is quickly evident when you look through the fines, whereby the only entity that wasn't a major company with hundreds of millions in turnover to break a fine of €5 million was a Croatian Debt Collector with absolutely appalling violations of basic data control - including processing minors, processing people with no debt at all, and monitoring things down to progression of terminal illnesses.
The most common by far is Art. 5 and 6 - Insufficient legal basis for data processing, followed by Art. 28 (3) and Art. 32 - Insufficient technical and organisational measures to ensure information security.
These are basic compliance requirements, mirroring something like PCI but for personal as opposed to cardholder information. Framing this as some lobbyist wet dream of Goliath vs David is just so much FUD.
The UI is kind of horrendous. Not even a dark mode. I also have a hard time imagining wanting to use BedRock at this point. Still can't use any of the leading models besides Claude, plus IAM config is comparatively painful, and you have to setup bedrock-access-gateway to get an OpenAI compatible API for tools that expect that API setup.
Amazon likes to force their devs into full stack roles, like they'll hire you for a frontend role and you'll end up writing java. Some teams do recognize the need for specialization, but generally speaking UI at amazon is an afterthought.
I asked it to summarize this thread [0] and it just said "that thread is about the monthly "who's hiring"' on HN", not even close.
[0] https://news.ycombinator.com/item?id=43533516