Giulio Ardoino

3.5K posts

Giulio Ardoino banner
Giulio Ardoino

Giulio Ardoino

@giuliastro

Innovation Manager, Crowdsourcer, tech Startups advisor, Accessibility and Blockchain enthusiast, founder of ventures such as AccessiBit, BestCreativity

Chiavari, Liguria Katılım Ekim 2007
307 Takip Edilen712 Takipçiler
Giulio Ardoino
Giulio Ardoino@giuliastro·
@AshClashYT Random Mortis in BrawlBall. Mortis in your team = you automatically lost the game.
English
0
0
0
17
Code: AshBS
Code: AshBS@AshClashYT·
Tell me YOUR biggest problem in Brawl Stars. 👇 I’m going to fix EVERY problem for Supercell. #BrawlStars
Code: AshBS tweet media
English
408
40
1.4K
81.8K
antirez
antirez@antirez·
If the correctness tests will pass, expect a sensibly faster DS4 inference speed for DGX Spark, and especially a lot flatter prefill as context increases. Soon in the repo if everything goes as expected.
English
14
6
197
24.4K
Giulio Ardoino retweetledi
Ezlivi
Ezlivi@ezlivi·
I played brawl ball for an hour and missed like 5 goals that I don’t ever miss. This is 100% new to the game and 100% buns
English
68
63
2.6K
197.2K
Giulio Ardoino
Giulio Ardoino@giuliastro·
@Pranit I wouldn’t risk too much when Open Source models are improving this fast. The gap between coding well and coding better will eventually shrink. And if an Open Source model is “good enough” at 1/15 of the cost… that changes everything.
English
1
0
0
118
Pranit
Pranit@Pranit·
OpenAI didn’t launch a new tier. They launched an upsell funnel. Plus users were paying $20 with Codex included. Now OpenAI is “rebalancing” that usage (read: reducing it) and introducing a $100/mo tier with 10x Codex access through May 31. The timing isn’t a coincidence. You reduce what people get at $20 at the exact moment you offer 10x at $100. The gap between the two tiers does the selling for you. At the end of May, the 10x drops to 5x. But by then you’ve already upgraded and $100/mo is just what you pay now. Anthropic did the same thing with Claude. Limits quietly got worse, then a temporary 2x off-peak promo showed up to ease the sting. Every AI subscription is subsidized at a loss. This is what the correction looks like: not a price hike, just a slow squeeze until the next tier up feels inevitable.
OpenAI@OpenAI

We’re updating our ChatGPT Pro and Plus subscriptions to better support the growing use of Codex. We’re introducing a new $100/month Pro tier. This new tier offers 5x more Codex usage than Plus and is best for longer, high-effort Codex sessions. In ChatGPT, this new Pro tier still offers access to all Pro features, including the exclusive Pro model and unlimited access to Instant and Thinking models. To celebrate the launch, we’re increasing Codex usage for a limited time through May 31st so that Pro $100 subscribers get up to 10x usage of ChatGPT Plus on Codex to build your most ambitious ideas.

English
94
47
812
126.3K
Giulio Ardoino
Giulio Ardoino@giuliastro·
@tonysimons_ @JoelDeTeves Hermes OpenViking plugin makes OpenViking not completely reliable. My agent had to patch the plugin multiple times, sent a PR with the fixes to Hermes. I have a local install of OpenViking with local models for embeddings and OpenRouter for retrieval.
English
0
0
0
49
Tony Simons
Tony Simons@tonysimons_·
@JoelDeTeves I’ve been using OpenViking and haven’t had any problems with it yet. 100 free and locally hosted!
English
3
0
4
1.5K
Joel - coffee/acc
Joel - coffee/acc@JoelDeTeves·
POLL: What’s the best memory system for Hermes Agent and why do you prefer it? Note: X polls only support 4 options so please answer in the comments and RT for a bigger sample size! 1. Honcho 2. OpenViking 3. Mem0 4. Holographic 5. RetainDB 6. ByteRover 7. Other / custom
English
54
2
103
32.7K
Giulio Ardoino
Giulio Ardoino@giuliastro·
Anthropic analyzed 2M Claude chats. The takeaway: AI usage today looks more like collaboration than replacement. The real risk might be junior roles: if AI does the entry-level work, how will the next generation gain experience? Curious about your thoughts. #AI #FutureOfWork
English
1
0
1
33
Giulio Ardoino
Giulio Ardoino@giuliastro·
@billyrestey Nice! The only glitch I noticed is that he has a left hand instead of the right one when clicking the button. Apart from this it's amazing.
English
1
0
1
157
billy restey
billy restey@billyrestey·
i made this short film for $65 in 24 hrs
English
167
62
995
62.6K
Nicholas Mboya
Nicholas Mboya@niccomboya·
@giuliastro Very cool. Mobile-first control for dev infra is underrated. One idea: a “safe mode” view (read-only + quick status) vs “execute commands” view. Makes it easier to trust. Are you planning iOS or focusing Android first?
English
1
0
0
15
Giulio Ardoino
Giulio Ardoino@giuliastro·
@antirez @steipete Clawdbot = open Cloud Cowork / Cloud Code. Plugins/integrations/skills, persistent memory, WhatsApp & Telegram proactive every day assistant (cron-driven always awake). It's model-agnostic and massively vulnerable by design because people wire it into everything and expose it.
English
0
0
0
93
antirez
antirez@antirez·
@steipete That's nuts. I'm going to try it, but would like to understand better what it does. Is there any video available? I know I'm nearly 49, but the YouTube thing got me.
English
11
0
88
12.2K
Rork
Rork@rork·
Rork is the easiest way to use Claude Code with Opus 4.5 to build real mobile apps, publish to App Store in 3 clicks and start making revenue Reply to get a free 25$ subscription. Ends tomorrow!
Rork@rork

Introducing Rork 1.5 • Easy app monetization with RevenueCat • Built-in analytics (no Firebase needed) • The smartest agent based on Opus 4.5 & Claude Code • Rork Stars community where successful mobile app founders like @zach_yadegari (Cal AI), @alexsllater (QUITTR), @georgeLampro20 will help you grow your app & over 100 small improvements 👇

English
3.6K
178
2.8K
476.8K
Legacy (Fan)
Legacy (Fan)@LegacySiu·
Guess the player Level: Hardest 98% will fail
Legacy (Fan) tweet media
English
17.3K
276
12.9K
4.5M
Massimo
Massimo@Rainmaker1973·
A remedy against insomnia [🎞️ the.elizara]
English
10
53
465
60.6K
Giulio Ardoino
Giulio Ardoino@giuliastro·
@bobbyBS_ I just don't open star drops and keep them to be sure I have enough credits to unlock the new brawlers. I litterally have more than a hundred star drops to be opened every time and probably never will. I agree, just boring.
English
0
0
0
154
bobby
bobby@bobbyBS_·
I think something brawl desperately needs to change is the star drop system, when it just came out it was the peak of brawl popularity but now it seems so lame with the fixed rewards you can’t even get excited about it anymore
English
43
12
781
38.8K
Giulio Ardoino
Giulio Ardoino@giuliastro·
@cline You guys are unbelievable. Keep up the good work.
English
0
0
1
470
Cline
Cline@cline·
We believe open source models should work as well as proprietary ones in Cline. Here's what we did to make it happen: SYSTEM PROMPT: Reduced GLM-4.6's prompt from 56,499 to 24,111 characters. A 57% reduction. Faster responses, lower costs, higher success rates. PROVIDER FILTERING: Same model, different providers = wildly different results. Some endpoints completely broke tool calls. We now filter for quality (OpenRouter's :exacto saved GLM-4.6). The result: Open source models now achieve stable, production-ready performance in Cline. Not just "good enough" but actually competitive. We're open source because we believe the community knows their models best. Generic system prompts are holding your models back. The gap is closing, and we're here to help close it.
canvrno@canvrno

We discovered GLM-4.6 was failing in Cline not because the model was flawed, but because inference providers were silently corrupting it. Using the same weights across different endpoints produced completely different behaviors. The variance wasn't minor, it determined whether the model could function at all. Some providers emitted tool calls inside reasoning traces. Others hallucinated parameters. We were debugging the wrong layer of the stack entirely. The fix required three interventions: 1. Prompt reduction from 56,499 to 24,111 characters 2. Provider filtering for high-fidelity endpoints only 3. Workflow enforcement with strict sequencing to prevent premature edits OpenRouter's :exacto endpoint transformed GLM-4.6 from intermittently broken to production-stable overnight. This poses a material risk to open source AI: When users encounter provider-induced failures, they blame the model, not the infrastructure. Trust erodes. Open source suffers. Reliability must be a shared responsibility. Transparent reporting of quantization settings and behavioral differences should become standard practice. Full technical analysis in the blog below, including our complete methodology and prompt optimizations. @cline cline.bot/blog/cline-our…

English
9
28
321
63.5K
Giulio Ardoino
Giulio Ardoino@giuliastro·
I’ve just got access to Comet, the new smart AI browser from @Perplexity_ai, and it’s absolutely impressive. I’ve got 5 invites to share. 👉If you’d like to try it, drop a comment telling me why you think an AI-powered browser could be a game changer. #AI #Perplexity
English
0
0
0
39