mason

567 posts

mason banner
mason

mason

@nosam2006

building Vital : health data into daily decisions documenting the process, failures and what AI is actually useful for

Irvine, Scotland Katılım Mayıs 2025
32 Takip Edilen15 Takipçiler
Sabitlenmiş Tweet
mason
mason@nosam2006·
I had never coded before, or never made a program that myself had used never mind other people. I wanted a health dashboard, deeper than the current health dashboards we have that actually sort through your data and give intelligent responses. I used Athlytic for a while, Bevel for another while but they didn't seem to give me the useful insights that I wanted. I went on Claude and coded an app, it looked like slop at first and was so basic. But I made it for me, and then I had the idea that I could share it with the world, that, maybe other people out there were missing that same piece of software. I know you may think this is an oversaturated market, and that apps do it better. But what I've been building the past few months started off as a dashboard, and had evolved into something more. It IS evolving into something personal which learns you with your data, checks in with you and lets you know when you should lay off, or go hard in the gym, amongst other things. I'm in the midst of a complete redesign, deepening every system and rebuilding systems from scratch, I'm truly building towards a personalized experience. I have nothing more to say really, not yet, however if you're interested you can download it at tryvital.app And in the next few months the deeper, richer app will be born, and released, I have high hopes for it. Let me know if any of you would like to help me test it, preferably fitness enthusiasts
English
0
0
0
31
mason
mason@nosam2006·
@argofowl No surprise. No one wants 3.6 flash. Any ideas why google are falling behind?
English
0
0
1
13
🥔🥔🥔
🥔🥔🥔@argofowl·
gemini 3.6 flash is absolutely terrible what a surprise
English
9
0
22
542
mason
mason@nosam2006·
To be honest this a banger. Very unique. They're innovating. Being able to go through and actually teach Claude how to do something on my computer is kinda groundbreaking. I like it, will help me out a lot with testing of my projects and apps, and certain specific research tasks I have to get it to do!
Claude@claudeai

New in Claude Cowork: teach Claude a skill. Record your screen while you do a task, talk through it as you go, and Claude turns it into a skill it can run again. Find it under Record a skill in the + menu of the Claude desktop app. Available on Pro, Max, and Team plans.

English
0
0
2
27
Chubby♨️
Chubby♨️@kimmonismus·
Its release day. Gemini 3.6 Flash - It's cheaper than Gemini 3.5 Flash ($7.50 output instead of $9.00 output). - It outperforms Gemini 3.1 Pro in almost every benchmark. It's being compared to GPT-5.6 Luna and Sonnet 5. Priced between these two models, it generally performs better. So, it's a solid mid-tier release.
Chubby♨️ tweet media
leo 🐾@synthwavedd

Gemini 3.6 Flash benchmarks are out, and it's... beaten by other models on code tasks, and is only really consistently SoTA on vision and context benchmarks. But hey, 3.1 Pro is now so old 3.6 Flash outperforms it across the board 😭

English
62
39
716
66K
mason
mason@nosam2006·
If testing passes, everything works fine and you can't break it then it's safe to assume it's alright. Obviously don't take everything it says at face value if you're not an expert in the field or knowledgeable but if you're using it to aid, and know what you're talking about (and what to look out for) then you can easily spot hallucinations. I very rarely spot wrong results. (Although tbf it did spit out outdated research yesterday which I caught). As long as you're cautious and not being an idiot you can spot it a mile away.
English
0
0
0
5
mason
mason@nosam2006·
@Koolkat6000 @kimmonismus Yeah it'll be good for that, but they have to compete on the frontier. I hope their Pro models are comparable to the frontier. Other models are more competitive at a lower price is all I'm saying.
English
1
0
1
18
S
S@Koolkat6000·
@nosam2006 @kimmonismus I think that's their point, enterprise API use. Heck I use 3 Flash for that. (3.5 Flash is shit compared to 3 And more expensive)
English
1
0
1
21
mason
mason@nosam2006·
I'd appreciate testers! DM or reply if you're interested! V2 includes deeper recovery intelligence, which takes richer cues from sleep, stress, fatigue, illness and soreness data. It includes adaptive training plans, which adapt according to your data and check-ins. Furthermore, it includes richer metabolic intelligence across the systems, with higher levels of predictive analysis with the data you provide. Everything is currently being weaved together to provide a seamless experience, something I admit V1 did not do very well. There's plenty of more features I've missed out, as I've been working on this for weeks now, but I'm really excited to share it.
mason@nosam2006

This is the new UI I've been working on. It is more intelligent, less messy and gives more insightful outputs at every turn. This is the main part of it, offering one number based on a multitude of algorithms which calculate it, taking data from every domain. It's very much a work in progress but you can see the tips and next steps it intelligently suggests based on the sleep you've had, nutrition from the previous week and the current week, metrics such as wrist temperature (which can be used to track illness) and other metrics such as activity level throughout the week and fatigue. You name it, we're trying to track it. Furthermore, at the bottom you can see the check in button. This is a new feature I've added in the past week or so, upon building the prototype, that lets the user (and forces them) to manually calibrate their score. It asks them a multitude of questions, which takes under a minute. This adds a second layer to the system, which combines data from a wearable and all data from Apple Health with manual, user entered data, which makes the final outcome all the more accurate. It's an early prototype, but in my opinion is much better than what we had before. The dashboard we had before did not feel alive, nor was it very insightful. It just showed data, and let the user infer, which was not the way to go. You live and you learn, and over the next few weeks I'll be sharing more as it evolves. I really hope these systems will help people who are struggling with their health across all domains, and as it gets better it will be better equipped to help. Again, we're very early. Vital V1 is on the app store right now. This version, which I'm coining Vital V2 will be on the app store late august/early september, and it will be a very, very meaningful improvement. I'll share more updates soon!

English
0
0
0
18
mason
mason@nosam2006·
Vital V1 is available now! tryvital.app If you want to test V2 before release (and trust me, I'm looking for testers so don't hesitate to reach out, please reply or dm me. I'm especially looking for people who train reguarly, are in tune with their body and use Apple Health! We can get something sorted. Thanks).
English
0
0
1
7
mason
mason@nosam2006·
This is the new UI I've been working on. It is more intelligent, less messy and gives more insightful outputs at every turn. This is the main part of it, offering one number based on a multitude of algorithms which calculate it, taking data from every domain. It's very much a work in progress but you can see the tips and next steps it intelligently suggests based on the sleep you've had, nutrition from the previous week and the current week, metrics such as wrist temperature (which can be used to track illness) and other metrics such as activity level throughout the week and fatigue. You name it, we're trying to track it. Furthermore, at the bottom you can see the check in button. This is a new feature I've added in the past week or so, upon building the prototype, that lets the user (and forces them) to manually calibrate their score. It asks them a multitude of questions, which takes under a minute. This adds a second layer to the system, which combines data from a wearable and all data from Apple Health with manual, user entered data, which makes the final outcome all the more accurate. It's an early prototype, but in my opinion is much better than what we had before. The dashboard we had before did not feel alive, nor was it very insightful. It just showed data, and let the user infer, which was not the way to go. You live and you learn, and over the next few weeks I'll be sharing more as it evolves. I really hope these systems will help people who are struggling with their health across all domains, and as it gets better it will be better equipped to help. Again, we're very early. Vital V1 is on the app store right now. This version, which I'm coining Vital V2 will be on the app store late august/early september, and it will be a very, very meaningful improvement. I'll share more updates soon!
mason tweet media
English
1
0
2
30
mason
mason@nosam2006·
3.6 flash lite may be v useful for me when it comes to Vital. Right now we use a mixture of Sonnet/Haiku with an autorouter depending on the task at hand. It decides what is needed based on the input, but if 3.6 flash operates at a good enough quality, with high enough speed and pricing then I might make the switch. Right now I'm all about optimizing my API costs, it looks promising. I guess we'll have to wait and see what happens. Google has always been the king at pricing to quality ratio.
🚨 AI News | TestingCatalog@testingcatalog

GOOGLE 🔥: Gemini 3.6 Flash and Gemini 3.5 Flash Lite models are now available on Google AI Studio and Vertex API. > gemini-3.6-flash: "Our most intelligent model yet for sustained frontier performance in agentic and coding tasks." > gemini-3.5-flash-lite: "High-throughput, low-latency execution for scaling high-volume agentic tasks and subagent workflows" Gemini? 👀

English
0
0
0
18
mason
mason@nosam2006·
Yeah no one wants 3.6 flash at all. Google have been silent for a while and then release a Sonnet level model, when we have frontier models. This will not help me in my projects, and I think a lot of others will feel the exact same. The only saving grace is if the quality/price/efficiency ratio is great, then it may be useful.
English
0
0
0
149
Harshith
Harshith@HarshithLucky3·
Gemini 3.6 flash is a good model but we want Gemini 3.5 pro 😭
English
7
2
60
12.9K
mason
mason@nosam2006·
@Da7_Tech Wouldn't they make more money by targeting the general audience?
English
1
0
0
8
Da7em
Da7em@Da7_Tech·
Mark my words: OpenAI will either phase out regular ChatGPT or heavily restrict its usage. Their real priority is that Work nonsense. Every time I give ChatGPT a task, it immediately pushes me toward Work instead. The direction couldn’t be clearer.
English
13
3
30
2.2K
mason
mason@nosam2006·
@SahilPanhotra What instances of Fable hallucinating have you seen? I've used Fable for countless hours building my projects and haven't encountered it. Would love to learn more though!
English
1
0
1
11
Oscar Wing
Oscar Wing@fried_ege·
@nosam2006 @scaling01 Read his Substack Chinese frontier is 6 months behind and will take 8-12 months to catch up to mythos
English
1
0
2
17
mason
mason@nosam2006·
That's what I realized when building my health app. Hell, Opus 4.5 could build you a working app, and now we have Fable 5 and Sol. The bottleneck is how much taste you have, how much you're willing to iterate on the product, take user feedback into account, and how long you're willing to stick with the project not giving up. The models getting better means nothing for the average person. Most things could've been built 6 months ago. It just allows the creative mind to build whatever they want. I'm very glad it's enabled me to bring my ideas to life.
English
0
0
0
27
Santiago
Santiago@svpino·
Coding is easier than ever, but thinking is still hard. Nothing has changed. The bottleneck is still your brain, not a model.
English
39
30
273
9.7K
Bill G.
Bill G.@billgnofficial·
Explain your product in "one sentence" below. If you can communicate its value clearly, I'll visit your website and give you genuine feedback.
English
73
0
28
1.9K
mason
mason@nosam2006·
Yeah that's where I am right now to be honest. When I started building my health app I was really optimistic, excited and had it in my mind it's going to work, but as time has went on my confidence has been shifted a little if I'm being honest. I know if I just keep going, iterating the systems within the app and making it a great experience for the user then it'll come. I just need to keep my head up. This is where most people quit though tbf so by recognizing that I'm already ahead. Hopefully I'm wrong!
English
1
0
1
4
Kiril
Kiril@BuildWithxAI·
Every app I've launched started the same way. "This probably won't work..." Funny how often that's wrong.
English
9
0
8
296
mason
mason@nosam2006·
Working on debloating my health app. Codex bloated it so bad with tests and over engineered code, so I've tasked Fable with identifying all of the bloat, and am slowly working through debloating it. Performance tanked after Codex finished implementation but I guess that's just how it is. Very, very eager model though. Perf is slowly getting better lol
English
0
0
0
7
mason
mason@nosam2006·
100 active users for sure. on my health app (when I start getting users) the sheer user data and feedback will help me tune it to give even more insightful responses and make the algorithm better. 10,000 visitors really doesn't do anything for me yk, no insightful or helpful data. all I'm missing atm is user data, that is my bottleneck currently.
English
0
0
1
4
Bowora
Bowora@bowora_com·
Would you rather have 100 active users or 10,000 visitors?
English
10
0
9
558

Keşfet

@argofowl @Koolkat6000 @kimmonismus @Da7_Tech @elonmusk @BarackObama @taylorswift13 @cristiano