Yusril Izza Aulia

Javascript Developer | QA Engineer

[Yusril Izza Aulia @underworld146]

Halo tmen tmen mutualanku, Btw aku bru buat aplikasi web buat confess secara anonim. Selengkapnya simak thread berikut ini >>>

[saturn @SatOnchain]

Google just dropped a free 2-hour Graph Engineering course.

Most people stop at:
→ 1 prompt
→ 1 agent
→ 1 workflow

This goes much further:

17:44 → Build your first agent
39:30 → Loop engineering
1:12:38 → Graph engineering
1:34:26 → Self-throttling agents
1:55:05 → Multi-agent graph systems

The future isn't a single agent.

It's networks of agents that coordinate, evaluate, remember, and improve themselves.

Watch the course, then read the article below

[Avid @Av1dlive]

Andrew Ng just dropped a free 1-hour course on building agentic knowledge graphs from scratch

Watch it today, then read the article below on how to become a graph engineer.

Save and bookmark this no matter what. It'll be the most productive thing you do this week

[↓ Save this playbook before it disappear in your feed]

• 00:00 : What agentic knowledge graphs are and why agents need them

• 03:07 : How to construct your first agentic graph

• 14:00 :How multi-agent systems get architected on top of graphs

• 23:00 : How to build agentic graphs with Google's ADK

• 01:06:03 : Why graphs are the future of agentic AI

[Machina @EXM7777]

https://t.co/Kp41msNeIg

[Robin Syihab @anvie]

Saya baru saja mencoba pendekatan baru dalam manage context memory pada Agentic AI, pendekatan ini memungkinkan agent tidak pernah kehilangan memory-nya sekaligus mengoptimasi penggunaan context-nya, tanpa RAG, tanpa retrain model.
Saya kasih nama CMP: Context Memory Path.

[Rahul @sairahul1]

Google just dropped a 1-hour course on agentic engineering from scratch:

00:00 – How to build your first AI agent
08:24 – Build agent memory (short, persistent, long)
28:34 – Agentic loops, long-running AI agents
40:04 – How to build MCP (MCP vs API)
1:00:22 – Multi-agentic systems

This 1-hour watch will replace 10 paid agentic courses on the internet.

Bookmark this. Watch this weekend.

[codila @0xCodila]

Ex-Google engineer just dropped 1-hour course: loops, self-improving AI, memory systems - from scratch:

00:00 - the self-building agent
03:01 - soul.md runs everything
30:16 - RAG memory: pull 20 messages, not 2,000
31:48 - the loop that knows when to stop
35:14 - find the bug, fix the prompt
50:22 - how Claude compresses your memory

1 hour of his guide beats any paid agent course

watch & bookmark - then read Karpathy's loop method below

[Yusril Izza Aulia @underworld146]

@anvie apakah modelnya akan di publish di huggingface ?

[Robin Syihab @anvie]

Kalau kamu pakai Evonic dan punya GPU, set EVONIC_SESSION_ARCHIVE=1
Ini akan mengarsip conversation mu menjadi raw dataset yg kemudian bisa dioptimasi pakai model besar lalu digunakan utk training model kecil, hasilnya kamu akan punya model pribadi yg kecil, native dan powerful

[Hanako @hanakoxbt]

Anthropic Engineer Andrej Karpathy:

"Everyone builds agents on a model they've never seen. I deleted the framework to look underneath."

In 23 minutes he strips the framework Claude runs on and rebuilds it in raw C with zero dependencies.

What Karpathy actually did:

step 1 → Strip every abstraction. No autograd, no compiler, just float arrays that hide nothing.

step 2 → 2,000 lines train a GPT from scratch on a potato, then 60 strangers push it 20% past PyTorch.

The framework was never the hard part. Understanding what it hides was.

Watch the talk, then save the breakdown below.

[Codez @0xCodez]

https://t.co/eJB3TgyJwV

[Codez @0xCodez]

Ex-Google engineer reveals how to build AI agent loops, harnesses, LLM ops, and evals in 19 minutes.

Trace → evaluate → diagnose → fix → ship → repeat.

That loop is how agents self-improve over time.

Agentic loops + harness + memory + evals - that’s the senior engineer stack.

This is better than $500 paid courses on the same topic, explained in under 20 minutes.

Watch it, then save the framework below.

[Robin Syihab @anvie]

Invested heavily in this, integrating Evonic memory with Evomem knowledge infrastructure. pretty impressive for a local model to sort everything out this neatly. I feel it's a bit over-engineered, but yeah, it works

[Robin Syihab @anvie]

Tested Ornith-1.0-9B, and its impressive for a model of that size. I don't believe this is just 9B!

[Robin Syihab @anvie]

Evomem, blazingly fast knowledge infrastructure for agentic era, inspired by gbrain and Obsidian.

Ini akan menggantikan sistem kb @evonic_ai

https://t.co/1Tg6GkOIr2

[Robin Syihab @anvie]

Katanya ini model ditraining pakai data CoT Fable 5 sebelum ditakedown, mari kita coba, benchmark result soon

https://t.co/bekenkkR1k

[Charm @charmcli]

Qwen 3.7 Max is now on Hyper.

We waited a week to make sure it met our standards for performance and zero data retention.

This one is impressive. Try it here:
⚡️ https://t.co/pTUfhEzwM7

[Alok @analogalok]

i just ran Google's brand new Unsloth Gemma4 12B dense GGUF on my RTX 4060 using llama.cpp + CUDA 13.2

21 tokens per second. on a budget consumer GPU. locally.

no API. no cloud. no subscription.

and the benchmarks are absolutely cooked

# first let's talk architecture because this is genuinely different

every multimodal model you've used has a frozen vision encoder + frozen audio encoder + LLM backbone glued together

Gemma 4 12B is different

it's a single decoder only transformer. that's it. vision? raw 48×48 pixel patches → one matmul → projected directly into the LLM

audio? raw 16kHz signal sliced into 40ms frames → linear projection → same LLM input space

no encoder tax. no latency penalty. no fragmented memory

to put the encoder savings in perspective:

old Gemma 4 26B approach:

- 550M param vision encoder (frozen)
- 300M param audio encoder (frozen)
- LLM backbone

Gemma 4 12B:

- 35M param vision embedder (a single matmul)
- no audio encoder at all
- LLM backbone handles EVERYTHING 550M → 35M for vision alone. that's a 15x reduction

this is why the gemma-4-12b-it-Q4_K_M.gguf is just 6.6 GBs!!!

and it has 256K native context context

# Benchmarks:

AIME 2026 (math olympiad): 77.5%
GPQA Diamond (expert science): 78.8% LiveCodeBench v6 (real code): 72%
Codeforces ELO: 1659
MMLU Pro: 77.2%
MATH-Vision: 79.7%
BigBench Extra Hard: 53%

inference → llama.cpp, LM Studio, vLLM, SGLang

llamacpp flags:

-m "gemma-4-12b-it-Q4_K_M.gguf" -ngl 99 -c 8000 -v --port 8080

Available on huggingface now! Link below

[Yusril Izza Aulia @underworld146]

@perogeremmer pakai plan yang apa mas ? aku pakai allgreto makin hari makin lemot

[Yusril Izza Aulia @underworld146]

ada yang jual macbook air m4 24/512gb area Jogja?. BNIB/Second boleh.

#macbookairm4 #jogja

[Ibrahim Arief @ibamarief]

Dari 4 hakim anggota, 2 hakim vonis Ibam 4 tahun, tapi 2 hakim lainnya justru minta suami saya dibebaskan.

Sampai dua hakim menyatakan keyakinan berbeda lewat dissenting opinion terhadap tiga hakim lainnya jarang sekali kejadian.

Bagi keluarga kami, ini tanda nyata kalau sebenarnya ada keraguan yang sangat mendalam dari majelis hakim sendiri soal perkara Ibam.

Rasanya sedih banget, kenapa putusan yang menentukan masa depan keluarga kami penuh keraguan seperti ini?

Bahkan di ruang peradilan, hakim ketuanya pun terlihat ragu-ragu untuk memutus berdasarkan pertimbangan vonis 4 tahun seperti 2 hakim anggota yang lain.

Padahal Bapak Presiden @prabowo selalu ingatkan kalau putusan penegakan hukum itu nggak boleh ada keragu-raguan sedikit pun.

Sedangkan, dua hakim anggota yang berkeyakinan Ibam secara terang benderang seharusnya dibebaskan dari seluruh dakwaan dalam putusan kemarin, dari matanya tampak kalau tidak ada keragu-raguan untuk menyatakan dengan penuh keyakinan bahwa Ibam seharusnya bebas.

Aku terharu banget, ternyata masih ada hakim yang sangat amanah dan objektif melihat kasus Ibam, terlebih ketika puluhan isi keyakinan kedua hakim yang mulia tersebut dibacakan di sidang.

Semuanya tentang Ibam yang aku kenal, semuanya berisi tentang keadilan yang muncul dari fakta persidangan yang dipahami secara utuh.

Tapi kenapa, putusan akhirnya penuh kejanggalan dan kezaliman? Ini sangat jauh dari keadilan yang didasarkan fakta persidangan.

Keyakinan kedua hakim tersebut bantu memperkuat keyakinan keluarga kami juga, kalau Ibam seharusnya bebas. Bikin kami makin yakin juga untuk terus berjuang demi keadilan.

Mohon terus bantu kawal ya, teman-teman.

Tolong bantu suarakan juga ke Komisi III DPR dan Pak Presiden Prabowo, supaya keraguan yang sangat kentara terlihat ini bisa berganti jadi kepastian hukum.

Jangan sampai ada lagi orang yang niatnya bantu negara dengan keahliannya malah dikriminalisasi.

Terima kasih banyak atas dukungannya selama ini. 🙏