I often miss my regularly hairstyle change days. My mom was strict and only voted for LONG NATURAL HAIR. I got a little rebel. At first she didn’t talk with me, but later realised it’s better I’ll play with hair and get bored. Yeah, now I am with natural brown hair back in my 26
1
92
I got part time position as a humanoid robotics engineer. I am..excited but nervous. This person wants robot to speak in our language which unsupported by the robot + make this robot lead a music orchestra..
105
After yesterday’s hackathon I think I am done with them for at lease next 6 months. My teammates generated a lot of code. Okay. But never checked. I spent 2 full days to read everything for contradictions and fixing everything.
3
89
Another, third teammate blindly added the main readme directly from LLM without reading at all..broken diagram, and with many lies.
I was so exhausted. And two persons decided to ghost the team, so in the last hours I needed to do demo and presentation
3
I said: let’s make our agent to extract data from .pdf .csv .txt .html
Teammates said ok. They BOTH separately wrote .md files where said the agent can work with variety of files.
I check the code and see “if file is not pdf, fail and exit”
63
Yesterday had my 2nd and 3rd interview at Huawei. Okay..that was cool but I was too nervous. I was interviewed overall by 4 people in one room
I don’t think I passed. Anyway I’ll wait till Friday. Still it’s honour for me to be noticed by them
70
Sofia retweeted
Meta published a paper that might end the current LLM era.
They proved that we can kill the "tokenizer" forever and it actually makes the models stronger.
Every AI you use today (ChatGPT, Claude, Gemini) reads text the exact same way. It chops your words into arbitrary chunks called tokens.
It relies on a massive, bloated vocabulary of around 100,000 specific chunks just to understand a sentence.
We assumed this was the only way to build AI. We treated it as the fundamental speed limit of language models.
Meta proved we were completely wrong.
Their new paper, "Breaking the Token Ceiling," ditches tokens entirely.
Instead, they forced the AI to process raw bytes. A vocabulary of just 256 values.
The results rewrite the rules of model efficiency.
By training on pure bytes and distilling the knowledge, these tiny models didn't just survive. They dominated.
Meta's distilled 1B-parameter byte models asymptotically surpassed state-of-the-art token models like Llama 3.2-1B and Gemma-3-1B across downstream tasks.
They use a fraction of the vocabulary. They require less data. They take up less storage.
It forces the model to actually learn the fundamental relationships between raw data instead of relying on pre-packaged text chunks.
This is a massive shift for anyone building, training, or deploying AI.
We’ve spent the last five years obsessively optimizing how we chop up text to feed these models.
It turns out, the bottleneck wasn't the intelligence of the model.
It was the tokenizer.
I also applied for volunteering in projects several weeks ago with computer vision/machine learning, but they ghosted me.
When people reject me even for non-profit projects - it aches a lot
Chinese professors ignored me (I have 1 Q1 journal paper + 1 A core paper)
Japanese too
57
These days I am preparing for 3 more hackathons:
1 in AI agents
1 in ROS for humanoid robot
1 in programming
Through September I went through also 3 hackathons + non stop programming intensive. In the end I have +7 new projects (in which 3 are analytics with SQL)
40
I was at career day and went with printed CV. Two companies said they need someone like me, one person said he’ll ask about PhD.
result: all ignore me for a week already..I sent my CV directly to them and via their companies websites
It’s been a year since I can’t find job
29
One month - 3 hackathons. In 2/3 was a team lead
Phewwww tired. Did I go to rest? No, I went on-site test programming 14 days (without days off) intensive about Python, Data Analysis, SQL, AI agents, models evaluation
24
Sofia retweeted
Deja de postular en LinkedIn en 2027.
Deja de postular en InfoJobs en 2027.
Deja de postular en Upwork en 2027.
Pasé 4 años aplicando. 1 entrevista.
Mi esposa pasó 2 horas en estos 10 sitios. 6 ofertas de trabajo remoto al instante.
Usa estas 10 plataformas para encontrar trabajos remotos ↓
Sofia retweeted
ETH Zurich just open-sourced their entire 2026 robot learning course.
Not a MOOC. The actual course. Slides, lecture recordings, coding assignments, GitHub repo.
The curriculum goes from imitation learning and RL all the way to Vision-Language-Action models and foundation models for robotics.
Guest lectures from the co-founder of Physical Intelligence. The creator of Diffusion Policy. Pieter Abbeel. Dieter Fox.
12 weeks. Free. No signup.
Taught by Oier Mees and the team at ETH Zurich.
If you want to understand where robot intelligence is actually heading… this is the reading list the field is using right now.
📍[cvg.ethz.ch/lectures/Robot-L…]
——
Weekly robotics and AI insights.
Subscribe free: 22astronauts.com
Sofia retweeted
I'm hiring a SWE intern for Pinetree Research, and the salary is $300K
Before you ask: 300K annualized. And no, we don't want your resume, coverletter, or grades
Here's how to apply:
1. Fork the Solari repo: github.com/solari-sdk/solari…
2. Build a real use case with Solari (browsers, sandboxes, and/or desktops)
3. Publish it on a public Github account
4. Share it in a post on LinkedIn or X
Tag @harrychow_ and @getsolari
Use AI to build it. Seriously, we insist. We don’t care how you ship, we care that you can ship something great, and if you can ship if faster with AI, even better. (@im_roy_lee)
What happens next:
- We review every build that tags us
- If we like what you made, we schedule an interview
- If you're the right fit, we'll hire you on the spot
Know someone who can take on the challenge? Tag them below
For questions about Solari or the role, comment below or DM 🚀
May the best dev win
Also started to watch free courses about AI from Antropic. Now try every day to talk with LLM more useful. Provide always contex, make briefs after several messages to catch on, ask about delegation. Talk in English (English is my second language but my native eats many tokens)
19
Started to learn how to create agents and orchestrate multi agents
Now I am participating in hackathon Alpaca trading, using Groq for agents
In September I have two more hackathons: VLA in bimanual robot and ML RecSys+RAG (maybe +VLM)
16
Sofia retweeted
剑桥这回直接扔王炸了!!
AI & ML经典教材全集直接免费开放,PDF随便下。
想学机器学习又不想被割韭菜买高价课的,这十本刷完,底子基本就硬了。
顺序从易到难排好了:
1️⃣ 《机器学习理解》——理论算法一把抓,零基础入门首选
🔗 cs.huji.ac.il/~shais/Underst…
2️⃣ 《机器学习数学基础》——数学底子弱的先把这本补上
🔗 mml-book.github.io/book/mml-…
3️⃣ 《机器学习算法的数学分析》——深入数学原理
🔗 tongzhang-ml.org/lt-book/lt-…
4️⃣ 《深度学习理论原理》——搞懂DL背后的理论根基
🔗 arxiv.org/pdf/2106.10165
5️⃣ 《神经网络与机器学习》——神经网络的系统讲解
🔗 arxiv.org/pdf/1901.05639
6️⃣ 《图深度学习》——图神经网络入门必读
🔗 yaoma24.github.io/dlg_book/d…
7️⃣ 《机器学习的算法视角》——从算法角度重新理解ML
🔗 people.csail.mit.edu/moitra/…
8️⃣ 《概率论:理论与实例》——概率基础打牢
🔗 sites.math.duke.edu/~rtd/PTE…
9️⃣ 《应用概率基础》——概率论实战应用
🔗 sites.math.duke.edu/~rtd/EP4…
🔟 《高级数据分析》——数据科学进阶必备
🔗 stat.cmu.edu/~cshalizi/ADAfa…
说句实话,这些书没一本是轻松的,别指望躺着翻完。
但只要你能硬啃下来两三本,比听群里吹一年AI牛逼都管用。
































