bartman

At the discounted rates, upgrading from 3 Flash to 3.7 Flash is finally reasonable.

In my evals 3.6 Flash (pre price change) was usually a bit more token efficient than 3 Flash, so I‘m expecting same or even lower cost-per-task on 3.7.

Maybe a play by Google to deprecate 3 Flash soon.

The timeline OpenAI wants to present looks like this to me:

- Leaving employee (LE) needs a USB stick to copy data off their Mac to give to their Manager (MA) so someone at Apple still has access

- After LE left, MA has access to LE Mac and iCloud to still access some data

- At some point, LE asks MA to sign out off LE's iMessage account on that Mac as they suspect they'll get OpenAI-related messages soon and don't want Apple/MA to see them

- Eventually, MA deletes the Apple files in LE's iCloud

Wait until you want to convert the UG to a GmbH and realize that it's not simple or cheap at all.

It's not enough to have had properly filed tax returns every year, have a large enough profit-collection-line item in your books (25k EUR+) and then fill out a form.

No, if you want to use the profit your UG was required to accrue to raise your capital stock to 25k and rename it to a GmbH you need to get your annual accounts audited.

Or alternatively, you can pay in the difference between your current capital stock (e.g. 2k) and the 25k minimum for the UG and then rename the company and "just" have to pay for the notary, publishing to public records, court, ...

I've not tested their speech-to-text yet, but based on the docs it looks promising. Thanks for the suggestion!

Even in the commercial space, there’s a lack of production grade ASR APIs that support diarization and word level timestamps.

My experiences with Google’s Chirp have been horrendous, with it sometimes skipping sections of speech entirely, hallucinating speech where the audio contains noise, and unreliable word level timestamps. And this all is even with using their new audio prefiltering feature.

AWS works slightly better, but also has trouble with keeping word level timestamps in sync.

Whisper is nice but hallucinates regularly.

OpenAI’s new transcription models are delivering accurate output but do not support word level timestamps…

A lot of this could be worked around by sending the resulting transcripts through a few layers of post processing, but… I just want to pay for an API that is reliable and saves me from doing all that work.

I’ve found CIS Controls v8.1 to be good and sane, with actual benefits to security. Level 1 is a solid base, and Level 2 is good for picking from depending on where risks exist in your business.

CIS Benchmarks are worth a look too: They’re best practices for securing typical cloud platforms, SaaS and OS.

Can you write a bit more about your workflow? I've been thinking about doing the same, but since I'm very non-interior-design minded have struggled to ask the right things.

Like... What are your inputs to the model? Empty renders of the space, or more fully decorated views/ photos? Do you have a light harness around this to help you discover the style you like and then stay consistent with it?

Do you find that giving a lot of context around the space you're designing helps (it hasn't in my attempts)?

Have you by any chance tried this with GPT 4.1 too (also 1M context)?

Friends who live in Kreuzberg and Friedrichshain could see it pretty well. I'm a bit further south within Berlin where skies are minimally darker, but between 10pm and 11pm it was so bright that light pollution didn't matter.

Funnily enough, despite having lots of alerts set up it was my mom who texted me from northern Brandenburg as she spotted it after getting an alert from RegenRadar, of all apps...

We had intense aurora in Berlin, Germany. Green clouds dancing in the sky levels. Started around 22:10 local time or a bit earlier, and at this point there's only a faint red/green glow remaining.

Anecdotally, my APP3 have worked much better then expected when I worked with a drill and jigsaw this weekend. No weird noises and much more NC than the 2 had.

Might be worth trying different ear tips or asking for a replacement if software updates don’t improve it.

In Germany, regulations around asbestos are quite strict: You're not allowed to "seal it in" by e.g. putting floor leveling compounds on top of old asbestos-containing flooring. If there's asbestos found during renovations, it needs to be removed professionally.

When we re-did our kitchen we found asbestos-containing glue under a new-ish layer of tiles one of the previous owners of the apartment just laid on top. I wish regulations would already have been stricter back then (and that they would have been followed - another story...) as this surprise find caused massive delays to the construction and forced us to temporarily move out during the removal and decontamination.

One of the best ways to make DIY-ing in buildings built before 1994 (when asbestos was banned for construction in Germany) safe has been to buy a H-filtration class shop vac. It can filter out asbestos fibers and many other fine dust particles that aren't healthy to inhale and was barely more expensive than a comparably good vac.