jpalomaki

Teams/Zoom calls when not on desk. iPhone Max screen is just a bit too small when people are sharing their screen.

Depends on how much information and details you have. The agents.md always goes into context. Detailed testing or process information might be excessive, when agent is working on UI. Skills are pulled when needed.

It’s easy to end up with bunch of old stuff in AGENTS.md that latest models no longer need and what might just confuse them. See for example the story about Anthropic cutting 80% of their system prompt [1].

Maybe they feel just using the AGENTS.md written for another model and possibly different era gives bad user experience.

[1] https://claude.com/blog/the-new-rules-of-context-engineering...

Models are now so good, that "how the model feels" is getting more and more important. I think this is pretty interesting, because there's no clear metric for this that you can optimize for. It's a matter of taste, like good design.

"OpenAI will launch a portable, screen-free smart speaker as its first consumer hardware product, Bloomberg News has reported, days after Apple sued the AI start-up and two former employees of the iPhone maker for trade-secret theft." [1]

[1] https://techcentral.co.za/jony-ives-first-openai-device-an-a...

As human I also give wrong answers if if I know the right one. Sometimes I also give answers even when I don’t really know them.

When pushed, I then start thinking and realise my mistake. System 1 vs 2?

Session data lookup is one select to database that gives 0..1 rows and uses index. In most cases this is not something you need to worry about.

When many sites are collecting these photos, it increases possibility of them leaking. Since these are also used for KYC process in crypto sites etc, this in turn increases risk of identity theft.

Isn't training material the biggest problem for truly open source LLMs (such that could compete with top tier models)? The computation part can be solved with money, but compiling a comprehensive training set that could be freely shared and free of copyright issues is pretty much impossible.

Do we need to bring Keybase[1] "back"? The original idea, mapping your social media presence to certain encryption keys.

In the future it will be increasingly difficult to prove in online context that you are not a bot. Being able to show that your social media (HN, GitHub, etc) presence goes way back would be an option.

[1] https://en.wikipedia.org/wiki/Keybase

"yes, there were regressions in some use cases of rsync in the 3.4.3 release. I quite deliberately tried to err on the side of fixing security issues for that release, and there were some valid (but unusual) use cases that got caught up in the changes"

It’s sometimes convenient if database is the only ”stateful” component in architecture.

Also if all the "state" is in one database, then you have better chance of getting consistent backups.

Not sure if this is the future I want, but I've always thought the main idea of smart glasses is to automatically bring up information that is relevant in your current context. One part of this is to recognize who you are staring at.

"By integrating Ookla’s data products, including Speedtest®, Downdetector®, Ekahau®, and RootMetrics®, Accenture will help Communications Service Providers (CSPs), hyperscalers, and enterprises optimize the mission-critical Wi-Fi and 5G networks that power their digital core. [...] Ookla’s data platform is anchored by more than 250 million consumer-initiated tests per month, complemented by controlled drive, walk, and embedded testing options"[1]

[1] https://newsroom.accenture.com/news/2026/accenture-to-acquir...

How about the payments, what is the easiest way from corporate customer perspective in DACH? Let's say for smaller subscriptions, <1000€/month?

"Only $99 to add 5G Backup to any UniFI network" and "Fully unlocked for any compatible carrier with SIM and eSIM support". Wonder if there's some catch?

Tell it to create html summaries with diagrams and sidebar for navigation.

Or ask Codex to create image that explains xyz.

It’s likely because the quick thought is that auth is just user table with hashed password.

Then when you really start thinking about it, the list of requirements grows.

Of course it’s still totally doable for an average developer, but takes time and mistakes can be catastrophic. And maybe the time is better spent developing stuff that differentiates you from others.

Kubernetes sounds like overkill, but I've been running microk8s for few standalone servers. This feels a pretty good match when working with agents. Codex can manage the cluster also over ssh, schedule new pods, check statuses, logs etc.

”Engineers were actively encouraged to use tools like Claude Code and Cursor, even ranking them on internal leaderboards based on usage”

Token maxxing? Might explain high costs if you are actively encouraging developers to spend as much tokens as possible.

I don’t think people want that, but they are willing to accept that in order to get stuff done.

Why not replace the SMTP with an API and explicit permissions. When registering for a newsletter, I would explicitly grant the sender right to push stuff to my inbox. At any point I could revoke this right and the sender would get clear error message when pushing.

Old fashioned person-to-person email would work as it does. This would only apply to the app-to-user stuff, which in my case makes up >99% of my emails.

Put proper LLM into Siri. Encourage developers to expose the functionality of their apps as functions, allow Siri LLM to access those (and sprinkle some magic security dust over it).

Boom, you have an agent in the phone capable of doing all the stuff you can do with the apps. Which means pretty much everything in our life.

Sometimes easy performance trick is to split the CTE to separate queries, put the results to unlogged temporary tables and add whatever indexes the next step needs.

Obviously makes only sense for stuff like analytical queries that are not running constantly.

Somebody took a deeper look at Claude Code and claims to find evidence of Anthropic's PaaS offering [1]. There's certainly money to be made by offering a nice platform where "citizen developers" can push their code.

From Astral the (fast) linter and type checker are pretty useful companions for agentic development.

[1] https://x.com/AprilNEA/status/2034209430158619084

Home scale example with strawberries: https://www.youtube.com/watch?v=8LIhx0yoM7s

A good solution for memory would help with stickiness. But it's a hard thing to crack.

Both Anthropic and OpenAI are working hard to move away from being "just" the LLM provider on the background.

Facebook is lacking access to the interesting data. If you are in the Google ecosystem then your private and business life is likely already there.

Kagi is using Google search behind the scenes. I think that’s why it felt so easy to switch to.