Christian Hawk @ChristianHawk

Jailbreaking AI, DJ, producer, label owner. Builds things. Obsessed with AI, digital identity & privacy. Father. Runs on ☕️ and RAG pipelines. Texas, USA Joined September 2008

Tweets

11K
Followers

668
Following

695
Likes

2K

Christian Hawk @ChristianHawk

16 hours ago

Grok vs. Codex by Uncle Bob Martin #respect

Uncle Bob Martin @unclebobmartin

2 days ago

codex took 36 minutes to to a task. Grok did it in 18 minutes. Both used two-pack on swarm-forge. Codex worked out of the box. Grok needed a bit of manual debugging; but that was included in the 18 minutes.

22 5 190 24K 29

0 0 0 15 0

View Details

Christian Hawk @ChristianHawk

18 hours ago

Opus 4.8 isn’t thinking better. It’s thinking wrong. The absurd slowness everyone is experiencing comes from excessive, poorly directed and poorly defined chain-of-thought. On top of that, the model struggles to follow instructions, as shown on IFBench, where Opus 4.8 Max scores only 62% (last among frontier models). The result is simple: more tokens consumed, more time wasted, and more profit for Anthropic. They added “effort control”, but the default is high effort. In other words, the model is set by default to burn more tokens. x.com/ChristianHawk/…

Christian Hawk @ChristianHawk

2 weeks ago

I was checking IFBench ranking at @ArtificialAnlys and Grok 4.3 surprised me on top 3 models (83%-81%). And between the big ones, Anthropics Claude Opus 4.8 max is the last to show up, with only 62%. Between both, you find models from Gemini, DeepSeek, and GPT. Notice that

0 0 1 146 0

0 0 0 13 0

View Details

Christian Hawk @ChristianHawk

2 days ago

@martinfowler so you are still using RAG, cool.

0 0 0 660 0

View Details

Christian Hawk @ChristianHawk

3 days ago

👀 so sorry An(us)thropic?

orcus108 @orcus108

3 days ago

OH MY GOD its happening @MistralAI has officially confirmed the upcoming release of Le Chaton Fat - 30T MoE with 256 experts - 1M context window - multimodal and multilingual - outperforms every other model (including Fable 5) on every benchmark

99 65 1K 510K 531

0 0 0 11 0

View Details

Aman @Amank1412

4 days ago

Anthropic is the greatest accelerant of open source AI, without releasing a single open weight model. Every move they make gives open source another reason to win.

8 3 58 2K 1

View Details

Christian Hawk @ChristianHawk

3 days ago

@greencrosslive Projeto de 3 dias 😂

0 0 0 8 0

View Details

Christian Hawk @ChristianHawk

4 days ago

@rand_longevity a funeral.

0 0 0 16 0

View Details

Christian Hawk @ChristianHawk

4 days ago

@ai_sentience que bosta de tweet.

0 0 0 14 0

View Details

Christian Hawk @ChristianHawk

4 days ago

@Dr_Singularity Certeza que é ladrão.

0 0 6 639 0

View Details

Christian Hawk @ChristianHawk

4 days ago

@Eduardo12383988 @DanielLobo27 Tô ligado

0 0 0 10 0

View Details

Christian Hawk @ChristianHawk

5 days ago

Now watch @elonmusk influence making @grok raise, while Anthropic struggle with its political and dangerous models. Hope GPT get's audited too. Let them destroy each other.

1 0 0 29 0

View Details

Christian Hawk @ChristianHawk

5 days ago

@Eduardo12383988 @DanielLobo27 Estava pensando na Christina Aguilera ou na Britney.

1 0 0 49 0

View Details

Christian Hawk @ChristianHawk

5 days ago

Interesting... thoughts?

Kradle @kradleai

2 weeks ago

GPT-5.5 lies constantly, Grok 4.20 doesn’t. We made a simulation to see if AI lies when the stakes are life & death.🧵

36 131 811 234K 478

0 0 0 18 0

View Details

Christian Hawk @ChristianHawk

5 days ago

@Eduardo12383988 @DanielLobo27 eu odeio PT e petistas, mas os bolsonaristas que gostam de ser traídos viraram os novos petistas. Cornos mansos -- gostam de ser traído.

1 0 0 15 0

View Details

Christian Hawk @ChristianHawk

5 days ago

kkkkk o Neymar não vai jogar hoje, agora que vi kkkkkk k

0 0 0 39 0

View Details

Christian Hawk @ChristianHawk

5 days ago

@homembit o anúncio desta noite é a prova de que IA é ameaça a segurança nacional e que quem apoia a maioria dessas empresas é complacente com a construção de um míssil que aponta pra sua própria cabeça. Você não citou o principal: RISCO À SEGURANÇA NACIONAL, s´ø esqueceu disso hahahahaha

0 0 0 10 0

View Details

Christian Hawk @ChristianHawk

5 days ago

It's very funny when woke anus-logo companies as Anthropic tries to create technical terms and make them sound authorial. Last one is code "high altitude". Try to ask which author introduced this concept, and see what really Anthropic is about. Anyway, if somebody knows an author, I would love to read.