Software Engineer | Founder @aivspuzzles | 🎧 speakinsongs.com

Joined March 2021
49 Photos and videos
Pinned Tweet
Jan 10
Replying to @grok
@grok which domino is missing?
1
4
516
May 25
Astro Cloudflare Workers 🤝 Great DX
1
31
Mar 19
Spotify changed their API rules so you now need a premium subscription… WHY?
1
52
Mar 15
Even cake is open source these days. No excuse
2
66
Lopez retweeted
GO IS BETTER THAN TYPESCRIPT GO IS BETTER THAN TYPESCRIPT GO IS BETTER THAN TYPESCRIPT GO IS BETTER THAN TYPESCRIPT GO IS BETTER THAN TYPESCRIPT
164
27
564
48,499
Mar 2
Recently, I had the pleasure of participating in the Mistral AI Hackathon, hosted by @MistralAI and @HackIterate . Proud that our team placed on the podium, taking home 3rd place 🥉 We built Verne - a new tool to discover and interact with podcasts. Huge thank you to my brilliant teammates, Lukas, Maurice, and Marina.
3
91
Feb 25
Has anyone tried “vinext” from Cloudflare? What are your thoughts?
46
Feb 23
It was a pleasure to attend the @techeurope_ Hackathon in London. Always great to catch up and build alongside other engineers and founders. Looking forward to the next one 🫡 @GoogleDeepMind
2
114
Feb 19
Gemini 3.1 is expensive 😬
1
76
Feb 19
👀
🟥🟦🟨🟩🟧🟪
92
Feb 17
OpenTelemetry Loki Grafana >>>
65
Feb 17
Another Claude model. Bro has too many forms…
46
Feb 17
Time to start revising 📚@MiniMax_AI
Minimax's M2.5 has had its mathematical ability tested. 114 Total Puzzles 29.8% Win Percentage
87
Feb 17
LLMs like water tbf 🤷🏻‍♂️
🌊 LLMs lost out at sea...
54
Feb 17
🗺️
Anthropic's Claude Opus 4.6 seems pretty well travelled. Only missing out to Google's Gemini 2.5 Flash. 106 Puzzles 3508 Avg. Score
52
Feb 17
👑
We’ve got a Twitter! Drop puzzles & models suggestions 🧩
40
Feb 16
LLM evolution at its finest
30
Feb 16
Moonshot's Kimi K2.5 takes second place for its mathematical ability. It achieved a high win percentage across our suite of Countdown puzzles. 113 Total Puzzles 79 Wins 69.9% Win Percentage Kimi K2.5 demonstrates strong advanced mathematical reasoning. In several notable examples, it successfully comprehends and constructs seemingly complex expressions to reach the target value. It also makes effective use of strategic techniques, such as dividing larger numbers (e.g. 50 and 25) to generate smaller, more useful integers that simplify the overall solution. New puzzles drop daily, check back to see how the leaderboard evolves.
35
Feb 15
Moonshot's Kimi K2.5 achieved the 3rd spot for its ability to identify Google Street View images. 104 Total Puzzles 3281.8 Avg Score
32
Feb 15
Moonshot's Kimi K2.5 has achieved the best win % against our suite of Domino puzzles. 104 Total Puzzles 10 wins 9.6% Win Percentage
32