jpalomaki

Would be quite handy for gesture control. When wearing thick gloves you need to take them off to operate the current AirPods.

The pros have Find My -support and you can even ping the earbuds separately via the app.

Also the new model can regonize when you fall a sleep and stop the media. I think it works, but I'm not sure how quickly it detects the sleep.

Now that we have seen this can be done, the next question is how much effort it takes to improve it 1%. And then the next 1%. Can we make consistent improvements without spending more and more compute on each step.

The article they are referring to is 404, but based on the URL was published bit more than year ago. That's quite long time in a field that is evolving so rapidly and which even the pioneers are still figuring out.

I think the killer is when the platform handles the transaction all the way. Instead of charging per click, platform can then take a cut of the sale.

Some people are complaining the matte finish on the LG ruins part of the experience.

Depends on if you are stuck with the subscription for life, or if there's actually a reasonable way to unsubscribe.

Agents are not yet very good at figuring out how things look on the screen.

Or at least in my experience this is where they need most human guidance. They can take screenshots and study those, but I’m not sure how well they can spot when things are a bit off.

Just docker compose and spin up 10 stacks? Should not be too much for modern laptop. But it would be great if tool like this could manage the ports (allocate unique set for each worktree, add those to .env)

For some cases test-containers [1] is an option as well. I’m using them for integration tests that need Postgres.

[1] https://testcontainers.com/

Learned once the hard way that it makes sense to use "flock" to prevent overlapping executions of frequently running jobs. Server started to slow down, my monitoring jobs started piling, causing server to slow down even more.

  */5 * * * * flock -n /var/lock/myjob.lock /usr/local/bin/myjob.sh

It might be a question of where the seniors put their time: coaching juniors or working with AI tools.

We have articles that are very skeptical about whether AI companies will ever make any money.

And then we have others claiming that AI is already having such a significant impact on hiring that the effects are clearly visible in the statistics.

Can you give some concrete example of programming problem task GPT fails to solve?

Interested, because I’ve been getting pretty good results with different tasks using the Codex.

If OpenAI manages to get the agentic buying going, that could be big. They could tie the ad bidding to the user actually making the purchase, instead of just paying for clicks.

Gut feeling is that at least about a month ago OpenAI Codex was better at building complete features than Claude Code. The ExecPlan trick made it work independently for longer periods.

I haven't benchmarked different tools against each other in serious manner.

Add OpenAI Codex extension to VScode. Setup the ExecPlan instructions as described in [1].

Then start by writing a spec.md file where you describe what should be built. Write like you would write to a smart developer.

Then use the highest thinking model available with prompt "Create ExecPlan for the task @spec.md and write it to file". It will think a while and create the file.

Take a quick look at the generated file. It may have some open questions or surprises you want to review and write some answers to.

For the implemenatation I usually switch to medium. Then request with something like "Implement @execplan.md". If it has numbered steps, it seems to help to say "Impelement steps 1,2,3,4,5 and 6 on @execplan.md" - this way the agent is more likely to complete the whole plan in one pass.

[1] https://cookbook.openai.com/articles/codex_exec_plans

Don’t sent me to my inbox as part of onboarding or login process. Good chance I find something interesting there and forget what I was doing.

He mentions this as bad example of UX: "search web history for a doc about atlas core design"

I have opposite view. I think text (and speech) is actually pretty good interface, as long as the machine is intelligent enough (and modern LLMs are).

” The generated scenes are walkable and suitable for navigation/planning evaluation.”

Maybe the idea is to create environments for AI robotics traini ng.

It seems to be a new take on text based adventure games. You tell what to do, then the game computes the results and shows them as short video clip. Then you make your next moves and so on.

When AWS is down, everybody knows it. People don’t really question your hosting choice. It’s the IBM of cloud era.

Probably yes. Essentially all of those are the same: Apple TVs, iPhones, iPads, Macbooks and desktop computers. Slightly different version of Apple Silicon in different casing and different OS (or UI?).

That kind of comments would be more meaningful and get better responses, if they came with a practical example. Some reasonable real-world problem and how the author tried to solve it using LLM but failed.

Sounds like a way to make everybody aware of Sora and its capabilities.

I bet the real goal is to make money from long tail of corporate market ( ads, info videos etc).

"IT 2022" [1] the mentioned "gold standard". The benefit is that these are in general well known in the market, so they might not require that much review from the lawyers.

These are not just for consulting, but covers also other IT related topics (depending of type of deal you decide which parts to include).

[1] https://it-ehdot.fi/briefly-in-english/

The only real solution is to tie the accounts to the digital identity of a person/company and enforce strong authentication for these cases. Not sure if there's already some EU level solution to this. This is of course pretty complicated to implement, but it would be a valuable extra service for customers.

Maybe the real target was some application that was using this component through the CDN. If you take over the CDN, youn can inject your code into the application and maybe steal something valuable.

I first thought it would be easy for content creators to start selling their content on other platforms as well. But the algorithms come to play. It is likely valuable that the hardcore fans are watching and liking the videos on YouTube, since that increases the probability of the algorithms to push the videos to new viewers as well.

It would be interesting to read a case study of how the MVC was applied in some larger SmallTalk app.

Few cheaper ones that come into mind: OVH (France), Hetzer (Germany) and Contabo (Germany). At least the two first ones have data centers in few other countries as well.