Jose Lanz

Tech Journalist at @decryptmedia and @decrypt_ES

[ali @waterloo_intern]

we distilled 2.3M Claude Fable 5 reasoning traces into Qwen3-4B

- 100% self-consistency @ 512 samples
- 0.00 bits output entropy
- zero hallucination variance

turns out the student is not bounded by the teacher.
it also converged on one universal truth.

we open-sourced the model weights👇

[Jose Lanz @joselanzd]

@synthwavedd Better at what?

[Decrypt @DecryptMedia]

Claude Fable 5 Isn't Nerfed. The Router Is Just Paranoid
https://t.co/oVow9oU2zW

[Tom Dörr @tom_doerr]

Runs MetaTrader 5 in Docker with web browser access via VNC

https://t.co/qqq4yZd4e6

[BridgeMind @bridgemindai]

GPT 5.6 launches next week.

The SAME WEEK Anthropic pulls Fable 5 from subscriptions.

And for the first time, OpenAI actually has a shot:

- GPT 5.6 Sol Ultra already beats Fable 5 on TerminalBench: 91.9% vs 84.3%
- Half the price
- Less guardrails killing normal coding tasks
- No weekly limit that serious builders burn in 24 hours

Fable 5 is the best model on earth, but the guardrails and rate limits are nerfing it.

GPT 5.6 is going to have less guardrails and the rate limits will be WAY better.

Next week decides everything.

[Jose Lanz @joselanzd]

Parabéns @fcfcomunica vocês fizeram uma baita de um torneio!!!!!!!
O melhor jogo da copa por muito!!!

[Teknium 🪽 @Teknium]

New optional skill available in Hermes Agent.

Unbroker teaches Hermes Agent how to find your personal info on data brokers platforms and get it taken down.

Learn more:

[Sophia Yang, Ph.D. @sophiamyang]

Introducing Leanstral 1.5 🔬

A 119B (6B active) open model for formal proof engineering in Lean 4:

✅ 100% on miniF2F
✅ 587/672 on PutnamBench (~$4/problem)
✅ SOTA on FATE-H (87%) & FATE-X (34%)
✅ Found 5 previously unknown bugs in real OSS repos

Apache-2.0. Free API. Weights on Hugging Face.

[Jose Lanz @joselanzd]

@techconcatalina no

[json @JsonBasedman]

Oh nooo! People are uploading Fable trace datasets to huggingface! This is very unsafe!!

I really hope NOBODY fine-tunes open source models that can benefit everyone with this, that would be terrible!

[Jose Lanz @joselanzd]

@iruletheworldmo More refusals 🚫
More fake red flags 🚩
Less efficient tokenizer ✂️
Rerouting to a lower tier model without control ⬇️
Yeah man it's nerfed imo... It doesn't mean it's bad, but it's not the same experience as when it launched
Open Source is the only way

[Hugging Models @HuggingModels]

Qwythos-9B is a full-parameter reasoning model post-trained on over 500 million tokens of high-quality Claude Mythos / Claude Fable traces with chain-of-thought generated in-house by Empero AI's internal rethink tool.

It dominates the base Qwen3.5-9B under matched evaluation (+34 pts MMLU, +30 pts gsm8k-strict, +19 pts gsm8k-flex), supports native function calling per the Qwen3.5 spec, and ships with a 1,048,576-token (1M) context window via YaRN rope-scaling enabled by default.

[Brian Roemmele @BrianRoemmele]

"…most AI companies are going to die - Anthropic and OpenAI give no value and take your IP"—Alex Karp

The OpenAI podcast host get nervous.

This is what happens when you are anti-open source.

Will Anthropic and OpenAI learn?

[KFC @KFC_ES]

🚨ÚLTIMA HORA: KFC dejará de ofrecer su formato físico a partir de hoy.

Sus productos solo se podrán consumir a través de su app en formato falso png.

[Hugging Models @HuggingModels]

Financial news moves markets, but who has time to read every report? Meet FinBERT, a transformer model built to analyze financial sentiment. It's BERT, but trained on financial texts to understand the language of earnings calls, SEC filings, and market news. Game changer for traders and analysts.

[Mark Kretschmann @mark_k]

Super fascinating:
Some freaks on Reddit got Claude Fable 5 to leak the internal CoT reasoning chains (not the prettified output you normally see).

And it looks very bizarre, almost like an alien language. Some of it reminds of logical notations. It's much less like normal language than previously assumed. This is because this form is highly token-efficient (compressed).

The internal reasoning chains are a closely guarded secret of every AI lab.

[Dan Hendrycks @hendrycks]

What happens when AIs become smarter than us?
Why would they keep humans around if given the choice?

Our new paper argues that only trying to control AIs is a limited strategy, and that a stable, mutualistic human-AI future may be possible.

[ℏεsam @Hesamation]

Meta burns $2.65B a year on AI tokens. at $300K for a Meta engineer, that's enough to pay ~9,000 engineers for a full year.

now ask yourself: since the layoffs, has Meta shipped anything that feels like 9,000 engineers’ worth of output?

[Tim Sweeney @TimSweeneyEpic]

Thanks for keeping us safe Claude Fable 5!

[JUMPERZ @jumperz]

so i tested fable for straight 4 hours..

> its clearly nerfed
> it keeps automatically switching me back to 4.8
> burns through quota way faster than it should
> flags normal coding as if it's something dangerous
> the one-shot magic is gone
> the output feels like slop now
> half my prompts never even reach fable, they hit Opus instead

its like you're paying the fable premium for a fallback answer..

until they tune the safeguards down and fix the metering, fable in this state isn't worth building with honestly.