SJ

Corporate strategist by day |AI startup dreamer by night | From boardroom to #DeepTech | 🇮🇳|One AI agent at a time

[FireHacker @thefirehacker]

Is there a prompt guide for Fable?.

Fable uses most of the quota in just few prompts and still feels nerfed.

I tried to use fable for a serious task like product analysis. It gave sharp analysis however looks like model is shy about tool calls. It doesn't want to collect a lot of information and I had to push it hard to do real analysis.
Analysis overall is sharper than Opus , however this feels like a nerfed model .
Is there prompt guide or direction how to use this model effectively?

[FireHacker @thefirehacker]

I built a small visualization layer on top of a local Qwen3 in Pure C to understand LLM output

Image shows why sampling is not greedy decoding: a lower-probability token can still get selected when temperature/top-p keep it inside the candidate pool.
I would also love feedback on what would make a visualization like this more useful for learning:

- KV cache view?
- attention heatmaps?
- speculative decoding comparison?
- greedy vs top-p side-by-side?

[FireHacker @thefirehacker]

Claude Cowork with blender is so much fun, still work in progress will post the final scene soon.

Trying out if it can build basic geometry nodes scene like waves hitting a beach 🌊🏖️

[Fetchlens.ai @fetchlens]

52% of MCP servers are dead within 90 days.
But the median server has 6 commits — lifetime.

The protocol works. The logic layer doesn't exist.

Content goes stale. Tools stay isolated. Nobody monitors what fails.

Full research: https://t.co/xCk7HPZbce

[FireHacker @thefirehacker]

We are in “Buy an iphone or train your own model” era.
Full ultra lean 1B model trained for one day at a budget of 💰1000🫡

[FireHacker @thefirehacker]

New UI Preview feature on Claude Code is really great.

I gave it a screenshot and asked it to make a navbar prettier.
Instead of immediately editing CSS, it first asked me to choose a direction:
Refined gold pill
Sparkle prefix
Glow halo around text
That is the part I found useful.
For frontend work, “make it prettier” is not a coding instruction. It is a taste decision.
Claude Code did not jump straight from prompt to diff. It stopped at the subjective layer first.
The flow felt like:
visual context → design options → human choice → code edit All in a single clean flow.

[FireHacker @thefirehacker]

This MTP pull request merge is getting more attention than many model drops.

I first noticed MTP while looking at Qwen3.5-0.8B, and now llama.cpp support makes the whole thing more interesting.

My current understanding is that MTP mainly improves token generation, not prompt processing.

So it helps when the model is writing a lot:

chat, coding, long answers, agents, synthetic data, local assistants.

But if the workload is mostly huge prompt + short answer, then prompt processing is still the bottleneck.

People are mentioning around 1.5x to 1.8x faster token generation in some setups.

My question is: how useful is this overall in real local AI workflows?

Is MTP going to matter mainly for long generation and agent loops, or will it become a default feature people expect in models?

[FireHacker @thefirehacker]

This is one of the most crucial lessons in First Break AI.

It teaches you how to navigate @huggingface like a pro.

Not just:

download model → run notebook → move on

In this lesson, we go deeper.

We look at how open model repos are structured, how to read model files, how config.json connects to the actual model class, and how to trace from a Hugging Face model page into the Transformers code that runs the model.

We use Qwen3-0.6B as the learning model.

We also look at why Markdown matters so much in AI workflows: model cards, GitHub issues, README files, Discord, Cursor, Claude Code, planning docs, and AI-assisted work.

Then comes the biggest win: datasets.

Working with datasets is a core AI engineering skill.

I show 3 ways to analyze datasets on Hugging Face:

Croissant endpoint
Data Studio / browser viewer
load_dataset with Python, pandas, and plots

We inspect dataset structure, categories, response lengths, distribution, short examples, long examples, and how to think about dataset quality before using it for training or fine-tuning.

And this sets up the next part:

running Qwen3 directly in C, without treating Transformers as magic.

Lesson 01: Hugging Face Beyond Upload

Watch:
https://t.co/GF8ZCNk5WN
Free cohort:
https://t.co/0H4qIVOpGj

[SJ @Quantum_Pandit]

@NsNitinsinha Demand should be either make WFH an option or pay to compensate for the inflation and petrol cost needed for travel.

[SJ @Quantum_Pandit]

@Govindtwtt Managing a remote team is a new skill. Boomers generally do not like to learn new stuffs unless threatened by organisation politics or given a “bigger picture”

[SJ @Quantum_Pandit]

@IshitaJoshi Because most of the Indian managers and leaders do not understand how to get the work done by their team in remote or hybrid settings up. They still follow traditional management style. Need for them to upskill to fit in now.

[SJ @Quantum_Pandit]

@ananyashasau Been on WFH mode long before COVID. Many functions can easily be WFH. Just like many meeting can be a email or sometime even a word on Teams. Except for core non tech IT role everything can be a WFH.

[SJ @Quantum_Pandit]

@DefiantDevii

[FireHacker @thefirehacker]

🚨 Major supply-chain attack: Mini Shai-Hulud is back

Reported impact:

• 170+ npm/PyPI packages affected
• 400+ malicious package versions
• 42 TanStack packages, including @tanstack/react-router
• UiPath packages
• Mistral AI SDK packages on npm/PyPI
• OpenSearch JS client
• Guardrails AI + others
This one targets developer + CI/CD secrets, so teams should check lockfiles, CI logs, npm/PyPI installs, and rotate exposed tokens.

[FireHacker @thefirehacker]

Where are small Models like Qwen3 0.6B and Qwen3.5 0.8B used ?
Huggingface shows 2.88 million downloads this month.

I can see 2.88 million downloads per month for small Qwen3.5 model. I tried using earlier model 0.6B in a deep resarch workflow and it was very difficult to get something done with this model .

Firstly they have a very surface level understanding of concepts. Poor Semantic understand means they can get confused about the topic or the task.

Json outputs are often broken . Adding a layer of checks on top took much of my time while working with these models.

Slow resposne. This one depends on a lot of factors and can actullay be improved , still slow response is a buzz kill most of the time

I am very curious how is the community using these models.

[FireHacker @thefirehacker]

I used to get irked by the rituals around me, whether social, religious, or community-based. I did not even value the rituals I had unintentionally created for myself.

The passage defines rituals as “time architecture.” We cannot inhabit time if its flow is not shaped or held by something.

Today, I see rituals as anchors. They are tools that help us center ourselves, calm the mind, and reduce anxiety, especially during turbulent times.

[SJ @Quantum_Pandit]

@coolfunnytshirt Question he wants us to ask is - what did he not eat… 🤔

[ア @yuruyurau]

t=0,d=5e-4,draw=_=>{t++||createCanvas(w=400,w);background(9).stroke(w,96);for(x=y=z=9,i=3e4;i--;point((q=x*(e=sin(t*PI/20-x*x/99+i%9)+1)+89)*cos(k=z/59-e/29+t*PI/480+i%9*8)+200,200-(q+60*cos(k/2))*sin(k)))[x,y,z]=[x+9*(y-x)*d,y+(x*(28-z)-y)*d,z+(x*y-z-z)*d]}//#つぶやきProcessing

[FireHacker @thefirehacker]

First Break AI
https://t.co/xLmTIU0rq6

Cohort: 1 May 2026 — 30 June 2026 (2 months)

3⃣ intuitions that make LLMs click:

🗿The model is a pipeline
For Qwen3-0.6B:
Input → text embeddings → Qwen3DecoderLayer ×28 → RMSNorm → lm_head → output

From far away, it looks simple.
Most of the intelligence is inside the repeated decoder layers.

🗿 LLMs generate one token at a time.
They are causal autoregressive models.
At inference time, the model sees the entire context so far, but it cannot see future tokens.
The loop is:
current context → predict next token → append token → new context → repeat
So the model does not produce the full answer in one shot.

It keeps extending the sequence one token at a time. Each new token becomes part of the context for the next prediction.

During training, the full sequence can be passed in at once, but a causal mask prevents each token from looking ahead.

🗿 The model does not directly output one word.
At every step, it outputs probabilities over the full vocabulary. For Qwen3-0.6B, that vocabulary is 151,936 possible tokens. Decoding then chooses the next token.

[Promita Mukherjee🇮🇳 @promzzz]

Calcutta cabbies stopping on their own and asking kothay jaben is not the poriborton I was hoping for TBH