Cesare Forelli

@cdf1982@iosdev.space
271 Followers
172 Following
5.6K Posts
Mac & iOS developer.
My apps: GlanceCam, Link HUB, PhotosUpload, Walk More, Tasktic, ContactsAMI, Always There and Weightrack
Websitehttps://cdf1982.com
GlanceCamhttps://www.glancecam.app
Milla's Instahttps://www.instagram.com/millakillapilla/
Cesare Forelli3m ago
Very unusual encounter during this morning’s walk
Cesare Forelli13h ago
Matt Corey15h ago

📢 Indie App Sales, July ’25 Edition starts TOMORROW 🚨

If you're an Indie Developer, but haven't registered yet, now's the time!

We have over 200 participants, but there's always room for more!

https://indieappsales.com

#IndieAppSales #AlmostTooLate

Cesare Forelli18h ago

Hello Australia?

I have so many questions about the 2.000 downloads of MarkCam in the last hour.

Clearly a MDM deployment of sorts, and sadly not one sale, but it’s still odd

Cesare Forelli19h ago

@matt1corey Please don’t hate me for the 7 entries in your Indie App Sales queue literally at the last moment possible! 😬

I’ve been meaning to add my apps for days, but everything’s been very chaotic.

I’m sorry to be a burden, but as always, thanks for doing this!

Cesare Forelli2d ago
Show threadJohn Gruber2d ago
@viticci It’s worse than that though. Doing something quickly, in reflexive reaction to the state of the competition, almost certainly won’t result in anything longterm. They need a competitive personal AI that’s the foundation for decades to come.
Cesare Forelli2d ago
Federico Viticci 2d ago

It's not just that Siri is "bad": the issue is that as people get used to basic LLM features, it increasingly feels like a product from 10 years ago.

Comparison of asking Siri to find the contents of a note Vs. Notion AI.

The 'Use Model' action in Shortcuts is a stopgap for power users, not something that most people can approach. Their Foundation models also feel like models from 2/3 years ago.

Apple needs to throw away Siri and replace it with an LLM with App Intents tool-calling ASAP.

Cesare Forelli2d ago

Proof that I love you all in my timeline is that when life was a mess like it’s been this week and I opened Tweetbot to 300 unread tweets, I just scrolled to top.

With you in Mona, I caught up. It took me a while, but I’ve read you.

Cesare Forelli2d ago
Sean Heber3d ago

ChatGPT and other AI services are basically killing @Iconfactory and I'm not exaggerating or being hyperbolical.

First Twitter/Elon killed our main app revenue that kept the lights on around here, then generative AI exploded to land a final blow to design revenue.

Cesare Forelli3d ago
Anuj AhoojaJun 26
In the last five years, we've gone from "employees will never have to go into an office" to "employees need to be in the office because creative and innovative work can only be done face-to-face between humans" to "lol we don't need humans"
Cesare Forelli5d ago
Vee6d ago
This is exactly what the internet is for.
×
Federico Viticci 2d ago

It's not just that Siri is "bad": the issue is that as people get used to basic LLM features, it increasingly feels like a product from 10 years ago.

Comparison of asking Siri to find the contents of a note Vs. Notion AI.

The 'Use Model' action in Shortcuts is a stopgap for power users, not something that most people can approach. Their Foundation models also feel like models from 2/3 years ago.

Apple needs to throw away Siri and replace it with an LLM with App Intents tool-calling ASAP.

Show threadgrauhausen2d ago
@viticci it does not feel like a product from 10 years ago, it is a product from 10 years ago that deteriorated substantially. Cannot even set my alarms within the first three tries in 2025. worked better years ago IIIRC.
Show threadDataless2d ago
@grauhausen @viticci I was gonna say the same thing. When Siri was launched they bragged about features like being able to tell it “my wife’s name is X” and having it know who you meant when you said “call my wife”. Stuff like that has disappeared. When I tell it to play a song now it tells me it can’t find it unless I state the exact title, performing artist, and which app to use, and even then it’s hit-or-miss. Sometimes it even launches the wrong app anyways. Used to be no problem.
Show threadDataless2d ago
@grauhausen @viticci That said, I don’t want an LLM bringing its own array of problems either, especially with current tech’s “always on everywhere” approach.
Show threadJeff2d ago
@viticci Yesterday I asked Siri (on iOS 26) “when was the iPhone 7 released” and it could only muster up a link to Apple's home page, that doesn't have the date that the iPhone 7 was released. Pretty weak!
Show threadEric Eggert2d ago

@viticci This is kinda comparing apples to oranges here. Of course notion will find notes in your notion notes. A search in the Notes app probably would also find the note.

Before “personal context” this is IMO an expected outcome.

(And just a simple word search would have worked, but why would we want to do anything without spicy autocomplete and a larger than necessary CO2 footprint these days?) (Sorry.)

Show threadNathan2d ago
@yatil But shouldn’t Siri be able to initiate a search in Apple Notes?
Show threadPavel Seleznev2d ago

@yatil @viticci This is iOS 17 (pre-Apple Intelligence).

I tried using voice search, and it seems that Siri can search the first line of the notes. Also without both cellular and Wi-Fi the search still worked.

Strangely, I had no luck searching the contents of the note. Siri's response was, “I didn't find anything.”

Long story short, it seems that pre-Apple Intelligence Siri indexed the headlines in the notes.

Show threadjoshhunt2d ago

@viticci I think even with some new LLM Siri, people would be severly disapointed in it. So often Siri would correctly recognise what I want to do (like turn lights on), but the tool-call will fail to execute the action.

Here's hoping that they really revamp how the whole system to make it actually work.

Show threadricharddas2d ago
@viticci This is not going to happen within any semblance of a useful timeframe, I’m afraid.
Show threadFurbland's Very Cool Mastodon™2d ago
@viticci alternatively, I will stick to my iPhone 6s which has NO “ai” and has a Siri that does what I tell it to when I tell it to do it without destroying the environment and ravaging the open web. thank you very much. /neg
Show threadFRANK.MCCONNEL2d ago
@GroupNebula563 I regret to tell you this may not help. TIme was I used to be able to say "Hey Siri, play This Kind Of Music" and get the Jonathan Richman track of that name which was in my in-phone music library. Later on newer phone I noticed it didn't work the same way, it just shuffle played my in-phone library. And indeed this change had also been made to Siri (via back-end service) on old phone too.
Show threadgok2d ago
@viticci what did models feel like 2/3 years ago?
Show threadJoe Lion2d ago
@viticci today in my car impressed the “ask Siri to play” button in CarPlay. She said “what would you like me to play?” I said “wake me up before you go-go”, to which the CarPlay Siri responded “what time would you like me to wake you up?” as if I had just said “hey Siri, wake me up…” But I didn’t say that - Siri SPECIFICALLY SAID “what would you like me to play?” I finally realized that the prompt that l CarPlay Siri asks me isn’t considered part of the conversation, like it would be for an LLM
Show threadDavid Cohen2d ago
@viticci Even ten years ago it wasn’t much good
Show threadJohn Gruber2d ago
@viticci It’s worse than that though. Doing something quickly, in reflexive reaction to the state of the competition, almost certainly won’t result in anything longterm. They need a competitive personal AI that’s the foundation for decades to come.
Show threadS★m V★rm★2d ago
@gruber @viticci IMHO this is /the/ play
Show threadTodd Heberlein (social)2d ago

@gruber @viticci I also think Apple should stay out of world knowledge in their models as much as possible. It is just too loaded with politics and religion. What we used to think of as a China problem (e.g., Tiananmen Square, flags for countries in Maps) is just as much a problem in the US now (e.g., 2020 election results, Gulf of America in Maps).

Outsourcing world "knowledge" lets someone else take the heat from an increasingly polarized population.

Show threadFederico Viticci 2d ago
@gruber Oh I don't disagree! Just very skeptical of what they can do realistically here for another year. Their new Foundation models are still a far cry from any other modern LLM. One has to wonder whether it's a matter of talent, lack of inference infrastructure, politics, or all of the above. Maybe new management can turn it around?
Show threadDuncan Babbage2d ago
@viticci @gruber the question I’m deeply interested to know is, do the new Foundation models make significantly fewer factual errors than any other modern LLM. That’s what instinct tells me Apple would like to do, even if it meant they moved more slowly. It could suggest a different approach being employed that might be a an advantage in the future. But that said, my confidence they are achieving this is low.
Show threadjaredrfrancis2d ago
@viticci @gruber I mean this a team that still has mastered search in their Mail app—an underreported poor user experience IMHO.
Show threadGreg Pierce2d ago

@viticci @gruber This may be implementation details, like providing the right "tools”, not a failing of the models themselves. I'm working on a tool that can query drafts with the Foundation Model and it handle this type of stuff pretty well.

You have to limit scope to avoid the 4k token limit – but still…

Show threadFederico Viticci 2d ago
@agiletortoise @gruber This is what I find most baffling. If it works reasonably well for developers, why isn't Apple doing this anywhere in its own apps?
Show threadGreg Pierce2d ago
@viticci @gruber It might be a scope problem. I imagine providing too many tools to the model might have diminishing returns, and they can't always make everything available (events, reminders, notes, etc.) – so I can imagine it is challenging to suss out what it applicable to a particular request.
Show threadHunter Hillegas2d ago
@agiletortoise @viticci @gruber You imagine correctly. At least the on device model turns to complete crap at tool calling after adding more than a few. Not surprising - GPT-4 Turbo (old but still much more powerful than AFM on device) couldn’t really handle more than four or five.
Show threadDrew McCormack1d ago
@hunter @agiletortoise @viticci @gruber Our experience too. We had hopes of doing some sort of automation with it, but have had to settle for a RAG setup that answers questions. It is basically summarizing search, and can handle that. It is way less powerful than even the cheapest OpenAI models. They probably should just outsource this stuff like they did search. Sounds like that is what they are planning.
Show threadJeff1d ago
@drewmccormack @hunter @agiletortoise @viticci @gruber I again asked Siri on my iPhone 16 Pro Max running the latest iOS 26 beta, “what year was the iPhone 7 released?” and this is the garbage response. Siri just isn’t capable of doing much at all this far on. They’d be much better off letting someone else do this stuff for them…
Show threadJeff1d ago
@drewmccormack @hunter @agiletortoise @viticci @gruber you would think that it would at least be able to give me an answer based on one of those two dates that it lists. Which one is correct? Personally, I am at least used to Siri giving me some kind of answer even if it’s wrong. In this case, Siri didn’t even try, just posted three links and said search Google yourself.
Show threadgok1d ago
@drewmccormack @hunter @agiletortoise @viticci @gruber I know you hear this a lot but please do file feedbacks with examples,
Show threadHunter Hillegas1d ago
@doctorgoktor Yup, filed a few.
Show threadgok13h ago
@hunter Thank you!
Show threadSimon1d ago

@hunter @agiletortoise @viticci @gruber
It sounds like they are at least a generation behind if they are performing at a similar level as pre 2024 models

https://www.understandingai.org/p/reinforcement-learning-explained

Reinforcement learning, explained with a minimum of math and jargon

To create reliable agents, AI companies had to go beyond predicting the next token.

Understanding AI
Show threadSteven Op de beeck2d ago
@viticci @gruber they run on-device instead of on a server with a terabyte of ram. Of course these baby-sized models are going to underperform the huge models.
Show threadRhys Morgan2d ago
@stevenodb @viticci @gruber I think this is including the on-server models running on Apple’s Private Cloud Compute machines.
Show threadSteven Op de beeck1d ago
@rhysmorgan @viticci @gruber is that already operational? I only see AI handing off to ChatHPT.
Show threadRhys Morgan14h ago
@stevenodb @viticci @gruber Yep, you can prompt it using Shortcuts. e.g. from Apple’s own “Morning Summary” example:
Show threadVille Turpeinen1d ago

@viticci @gruber other companies have basically given up on their “carbon negative by 2030” missions. Here’s how Microsoft is doing:
https://www.windowscentral.com/microsoft/microsoft-admits-that-its-carbon-emissions-have-soared-on-an-168-percent-glut-in-ai-energy-demand-we-recognize-that-we-must-also-bring-more-carbon-free-electricity-onto-the-grids

If Apple wants to offer ChatGPT level AI model for a billion users, they would probably need massive data centers that takes gigawatts of power, so they would also need to build gigawatts more wind or solar, if they want to stick to their 2030 plan:

https://www.apple.com/environment/

Microsoft sees its carbon emissions soar on an 168% glut in AI energy demand, "we recognize that we must also bring more carbon-free electricity onto the grids."

Microsoft's pledge to go carbon neutral by 2030 has hit an AI-shaped snag, as the firm's hunger for cloud growth impacts emissions targets.

Windows Central
Show threadMartín Pescador 🌍🐝1d ago

@viticci @gruber The one thing we know they have in big quantities is money. So they should have access to any resources required. So yes, for me it’s lack of good management and leadership.

And year after year they keep lagging behind the competition. It is becoming worse IMHO.

Show threadGabe Salkin1d ago
@viticci @gruber I keep floating that Apple needs to swallow their pride and buy a good model. @mgs suggested Mistral on an episode of The Talk Show and I still think that’s the best way to get a quality foundation model. Especially ones optimized for mobile and on-device
Show threadKrishna Permi2d ago

@gruber @viticci

Apple rarely used the word AI until last year, but here’s what Jeff Williams said seven years ago at TSMC 30th Anniversary Celebration Forum. It’s still astounding how they blew it.

From 9:20 https://youtu.be/ILV9X9Go3Hg?si=uJxjgQIe4WfQtF7F

Show threadJohn Gruber2d ago
@viticci But I completely agree that “Do I have a note about ____?” being something that normal people would (or if they’re outside the Apple ecosystem, already do) use regularly.
Show threadNate1d ago
@gruber @viticci asking for normal people on tech mastodon isn't a great sample, but, recently I've started asking Gemini to find my emails rather than searching Gmail. It's not every time but it slowly creeps up on you. My wife on the otherhand is much more 'normal' and vehemently dislikes talking to any assistant for anything.
Show threadRichfletcher21h ago
@gruber @viticci I’m not sure what it’s like stateside right now, but my daughter’s generation (she’s 13) are really not keen on Ai. Her peers here in the UK see it as a destruction of creativity.
Might be an ideological thing, who knows? Maybe Apple are just waiting that little bit longer to make sure they can say “I told you so” if/when an Ai pushback occurs.
Show threadLucian Marin2d ago
@viticci If Elon Musk could do Grok then Apple can do Siri LLM too.
Show threadFederico Viticci 2d ago
@lucianmarin Yup. I dislike Elon as much as anybody else but there's a lot of truth in there.
Show threadChris Gervais 🎸🥁💻1d ago
@viticci Does Notion use an on-device LLM or does it backend into a cloud inferencing pipeline? The on-device Apple Foundation models are tiny (not making excuses) and I hope with the betas they’re retraining or fine-tuning with each release to hone them further.