Phillip Fleischer
@plesioo.bsky.social
18 | Navigating my first software dev job while prepping for Abitur 🇩🇪. Sharing the reality of balancing code, studies & building. | Founder @ Young Devs Club: https://discord.gg/Eg4szKCf
👇Resources I use as a dev:
http://dev-resources.notion.site
👇Resources I use as a dev:
http://dev-resources.notion.site
This is my weekly schedule which balances work, Abitur prep, and side projects:
🧵 Let's break it down and turn it into 5 small tips at the end:
🧵 Let's break it down and turn it into 5 small tips at the end:
April 9, 2025 at 7:26 PM
This is my weekly schedule which balances work, Abitur prep, and side projects:
🧵 Let's break it down and turn it into 5 small tips at the end:
🧵 Let's break it down and turn it into 5 small tips at the end:
1/5 - "AI won't replace workers for now."
The transformer model don't has the potential to achieve AGI in their current form because it still lacks robust reasoning and data.
Just ask ChatGPT to create an image of a left hand and it won't work because of missing data, like here:
The transformer model don't has the potential to achieve AGI in their current form because it still lacks robust reasoning and data.
Just ask ChatGPT to create an image of a left hand and it won't work because of missing data, like here:
April 6, 2025 at 4:06 PM
1/5 - "AI won't replace workers for now."
The transformer model don't has the potential to achieve AGI in their current form because it still lacks robust reasoning and data.
Just ask ChatGPT to create an image of a left hand and it won't work because of missing data, like here:
The transformer model don't has the potential to achieve AGI in their current form because it still lacks robust reasoning and data.
Just ask ChatGPT to create an image of a left hand and it won't work because of missing data, like here:
Your brain tricks you about school work:
Realized again today. Thought studying would take 2 hours. Only took 30 minutes. Your brain tries to convince you to procrastinate because it dreads the effort.
But after finishing, it just feels SO good to have done it.
Realized again today. Thought studying would take 2 hours. Only took 30 minutes. Your brain tries to convince you to procrastinate because it dreads the effort.
But after finishing, it just feels SO good to have done it.
April 1, 2025 at 7:49 PM
Your brain tricks you about school work:
Realized again today. Thought studying would take 2 hours. Only took 30 minutes. Your brain tries to convince you to procrastinate because it dreads the effort.
But after finishing, it just feels SO good to have done it.
Realized again today. Thought studying would take 2 hours. Only took 30 minutes. Your brain tries to convince you to procrastinate because it dreads the effort.
But after finishing, it just feels SO good to have done it.
I wanted to quit german highschool to only code, but my mentor said no.
Now I'm so happy I followed his advice.
Here are the three reasons he told me to stay:
🧵 A thread ...
Now I'm so happy I followed his advice.
Here are the three reasons he told me to stay:
🧵 A thread ...
March 31, 2025 at 8:01 PM
I wanted to quit german highschool to only code, but my mentor said no.
Now I'm so happy I followed his advice.
Here are the three reasons he told me to stay:
🧵 A thread ...
Now I'm so happy I followed his advice.
Here are the three reasons he told me to stay:
🧵 A thread ...
But the future of AI is just speculation. Nobody knows what will happen tomorrow.
My teacher always told me the scene from "The Wolf of Wall Street."
Nobody knows! It's the same for AI:
My teacher always told me the scene from "The Wolf of Wall Street."
Nobody knows! It's the same for AI:
March 27, 2025 at 1:01 PM
But the future of AI is just speculation. Nobody knows what will happen tomorrow.
My teacher always told me the scene from "The Wolf of Wall Street."
Nobody knows! It's the same for AI:
My teacher always told me the scene from "The Wolf of Wall Street."
Nobody knows! It's the same for AI:
"AI will perform tasks month-long-projects in 2029" - METR Researchers last week reveal Moore's Law for AI agents
Every 7 months, AI's task endurance doubles.
This means AI does:
2026: 8-hour tasks
2028: Full workweek tasks
2029: Month-long projects
Every 7 months, AI's task endurance doubles.
This means AI does:
2026: 8-hour tasks
2028: Full workweek tasks
2029: Month-long projects
March 27, 2025 at 1:01 PM
"AI will perform tasks month-long-projects in 2029" - METR Researchers last week reveal Moore's Law for AI agents
Every 7 months, AI's task endurance doubles.
This means AI does:
2026: 8-hour tasks
2028: Full workweek tasks
2029: Month-long projects
Every 7 months, AI's task endurance doubles.
This means AI does:
2026: 8-hour tasks
2028: Full workweek tasks
2029: Month-long projects
10000 hours to become an expertat any skill? There is still a long way to go.
Consistency is key I guess.
What are your coding stats?
Consistency is key I guess.
What are your coding stats?
March 22, 2025 at 1:34 PM
10000 hours to become an expertat any skill? There is still a long way to go.
Consistency is key I guess.
What are your coding stats?
Consistency is key I guess.
What are your coding stats?
This is comedy gold. But shows the hard truth of "vibe coding" real saas
March 19, 2025 at 7:54 PM
This is comedy gold. But shows the hard truth of "vibe coding" real saas
The pricing difference between DeepSeek V3 and OpenAI GPT-4.5 is just insane. You basically get the same quality for approximately 1000 times cheaper.
February 28, 2025 at 6:29 PM
The pricing difference between DeepSeek V3 and OpenAI GPT-4.5 is just insane. You basically get the same quality for approximately 1000 times cheaper.
Claude Code is already full. Who secured a spot?
February 25, 2025 at 8:17 PM
Claude Code is already full. Who secured a spot?
Clearly, LLM models are moving towards solving tasks autonomously and collaborating effectively with teams. Looking ahead, Anthropic plans to not only allow AI to solve simple to medium tasks but also tackle tasks that "would have taken teams years to achieve".
February 24, 2025 at 7:31 PM
Clearly, LLM models are moving towards solving tasks autonomously and collaborating effectively with teams. Looking ahead, Anthropic plans to not only allow AI to solve simple to medium tasks but also tackle tasks that "would have taken teams years to achieve".
5/5 Most exciting: "Claude Code" - their new command line tool for agentic coding. It can search/read code, edit files, write/run tests, and commit to GitHub. Currently in limited preview, it automates complex dev tasks, completing 45-minute tasks in one go.
February 24, 2025 at 7:31 PM
5/5 Most exciting: "Claude Code" - their new command line tool for agentic coding. It can search/read code, edit files, write/run tests, and commit to GitHub. Currently in limited preview, it automates complex dev tasks, completing 45-minute tasks in one go.
4/5 New GitHub integration available on all Claude plans - connect your repos directly to Claude for better context on your codebase.
February 24, 2025 at 7:31 PM
4/5 New GitHub integration available on all Claude plans - connect your repos directly to Claude for better context on your codebase.
3/5 Major improvements in coding capabilities - especially front-end web dev. Benchmarks show SOTA performance on SWE-bench Verified and TAU-bench, with strong results across instruction following and reasoning tasks.
February 24, 2025 at 7:31 PM
3/5 Major improvements in coding capabilities - especially front-end web dev. Benchmarks show SOTA performance on SWE-bench Verified and TAU-bench, with strong results across instruction following and reasoning tasks.
Anthropic just released the best coding model to date: Claude 3.7 Sonnet and an agentic coding tool called Claude Code.
👇 Here is everything devs need to know:
👇 Here is everything devs need to know:
February 24, 2025 at 7:31 PM
Anthropic just released the best coding model to date: Claude 3.7 Sonnet and an agentic coding tool called Claude Code.
👇 Here is everything devs need to know:
👇 Here is everything devs need to know:
Grok 3 AI overthinks crazy. He spent 2 minutes on this request before I stopped him.
February 21, 2025 at 5:54 PM
Grok 3 AI overthinks crazy. He spent 2 minutes on this request before I stopped him.
Looks like Grok-3 is the new king of LLMs. Will you use it?
February 20, 2025 at 8:45 PM
Looks like Grok-3 is the new king of LLMs. Will you use it?
Recommended read for every person on Earth:
(Key takeaways in the next few days)
(Key takeaways in the next few days)
February 17, 2025 at 9:07 PM
Recommended read for every person on Earth:
(Key takeaways in the next few days)
(Key takeaways in the next few days)
At 18, I invested $290 monthly in a coding mentor, he gave me one piece of advice that stuck in my head.
February 6, 2025 at 8:07 PM
At 18, I invested $290 monthly in a coding mentor, he gave me one piece of advice that stuck in my head.