I would be really interested in real life throughput. For an agentic chat situation, we are still on 5.4 - not because of the cost, but it's simply much faster than 5.5 with comparable results. Also we are using gpt-5.4-mini a lot for quick summaries, tldrs etc.
In an ideal world we would upgrade 5.4 to 5.6 terra and 5.4 mini to 5.4 luna. But does somebody already have some measurements at least in terms of speed?
I would be really interested in real life throughput. For an agentic chat situation, we are still on 5.4 - not because of the cost, but it's simply much faster than 5.5 with comparable results. Also we are using gpt-5.4-mini a lot for quick summaries, tldrs etc.
In an ideal world we would upgrade 5.4 to 5.6 terra and 5.4 mini to 5.4 luna. But does somebody already have some measurements at least in terms of speed?
It took me a while until I found that you can switch between those two parts Work and Codex within the app on the left top corner when clicking on "ChatGPT Codex". And then - nothing happens. The old ChatGPT projects of my chats are gone, only appearing on the web. Wow, what a bad UX pattern.
If the EU just were to redirect the resources they're currently allocating to regulations like AI and Chat Control rather towards developing a genuinely competitive OpenAI or Anthropic alternative …
As much as I would like to agree with you about how non-digitized and rotten and crooked the whole notary etc system is - as someone who has incorporated and invested in dozens of German companies (UGs, GmbHs, and GmbH & Co. KGs), I’d seriously reconsider the UG & Co. KG structure for a one-person software business.
The tax explanation in the post is oversimplified.
Examples: With a KG, if the business earns €100k, that profit is attributed directly to you and taxed as personal income, whether you distribute it or not. At higher income levels, that’s roughly 45% including solidarity surcharge (and potentially church tax).
With a UG or GmbH, the company pays Körperschaftssteuer plus Gewerbesteuer, typically around 30% combined (depending on where its incorporated). On €100k profit, about €70k remains inside the company. If you later distribute it, you’ll pay capital gains tax based on the Halbeinkünfteverfahren on the distribution - which is 25% + Soli, bringing the total tax burden to roughly the same level as the KG.
The key difference is that with a UG/GmbH you can leave profits inside the company. That money can be reinvested into the business, other startups, ETFs, stocks, etc. - most often with only 1,5% effective tax while the money is working for you. You defer the second layer of taxation until you actually take the money out.
Also, you can pay yourself a salary. Whatever portion you pay out as salary is taxed personally just as it would be in the KG structure - but this time its company expenses, so no double taxation here either.
A few other points:
* A GmbH no longer requires €25k - its only 12,5k€ - and its also not to be locked away forever. The money can be used for legitimate business expenses immediately after incorporation.
* A decent tax advisor can usually get a VAT ID much faster than described in the article.
* A UG is widely accepted in the startup ecosystem. I’ve never seen customers reject an otherwise attractive startup because it started life as a UG. For investors, it's routine.
* Converting a UG into a GmbH later is routine, if you want to start small. If you have the 12,5k€ money, do a GmbH to save administrative hassle.
* A UG & Co. KG creates significantly more administration: two entities, two annual accounts, additional bookkeeping, additional filings, and additional advisor costs.
Of course there are valid reasons to use a GmbH & Co. KG, especially for complicated co-investment arrangements, but from what you wrote, that's not the case here. Therefore, for a solo software founder, I’d question whether the additional complexity buys you anything meaningful.
Bonus tip: You MIGHT want to consider owning your share in the Software UG not directly but through another UG, paolino UG or so. When you foresee to sell your business for significant money later, then you'll have exactly the same advantage, the money can stay in the company for reinvestment and you don't have to give up 50% of your capital gains in the moment you sign the sellers agreement. You CAN't do this easily later.
Not legal or tax advice. Just my personal experience.
You've hit your monthly spend limit.
/rate-limit-options
What do you want to do?
Adjust monthly spend limit: Unlimited ← or → to set a limit
Wait for limit to reset
I've never hit a usage limit on my Max plan, basically ever -despite heavy xhigh usage on Opus 4.8.
I added $133 credits which I still had from somewhere. That lasted 27 minutes.
I think we are being prepared for a Post-IPO-World in terms of pricing.
It's hard to take the author seriously technically competent when reading "DynamoDB what a hot pile of garbage. I tried it and ended up with a $75USD bill by the end of the day."
Clearly there are many people - including me - who built highly scalable, available and near maintenance-free systems using DynamoDB for a ridiculously low cost.
I have no idea how you can actually burn more than $5 in development for DDB. If you don't make the effort to explore what a technology is built for and/or clearly didn't understand it, maybe you should holding back ranting about it. Unless you want to look like a fool.
Same goes for IAM. It's complex but still easily understandable to get the basics. Creating e.g. a rule where you can only read from a DynamoDB table but not delete entries or the whole table takes you under 10 clicks.
I've been holding Hacker News in too high of esteem for too long I think. Most people on here are amateurs at best. The complaints about AWS/Cloud and LLMs are getting too loud. Did the industry pull in too many inexperienced during covid and now we're witnessing the backlash or something?
I've been doing this a long time. Running hardware, running VPS in the cloud, and running on AWS. I'll never go back to hardware/on-prem.
I have a pre-installed pool heat pump - an “AcquaSource” branded unit, the kind you can buy at any pool store in Europe - which supports WiFi. The App called “Pool Panel” wasn’t pretty, but it worked and I didn’t give it much thought. At one time, the remote control of the pump stopped responding: The pump itself was fine; the panel worked, the temperature held. So I decided to take a deeper look at how it all works. It turned out to be a security nightmare.
As icing on the cake: Their iOS app “Pool panel” by the developer “Guangzhou Wo Jie Information Technology Co., Ltd” is unmaintained since 2019, the contact link leads to a broken link (https://www.axen-heatpump.com/contactus.html), nobody responded via email. Very trustworthy.
This is the story of how I got control back, learned a few uncomfortable things along the way, and ended up with a small Docker container that exposes my pool pump as a clean local REST API.
