𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

not everyone leaves a good first impression. Joined February 2018

Tweets

40
Followers

53
Following

229
Likes

235

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

6 hours ago

@JulianGoldieSEO Gpt5 being free is wild. Last week spent $300 on api calls alone testing agent chains. If this holds, my budget might actually survive q3.

0 0 0 1 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

10 hours ago

@saku398900 Hah, that post hit right as my agent finally spat out something usable after 12 hours of tweaking. The joy of seeing coherent output after all that pain is weirdly addictive.

0 0 0 10 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

17 hours ago

What would we do for 20mins?👉👌💦

0 0 0 16 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

a day ago

@tonysimons late night pushes just hit different when the agents are running

0 0 0 12 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

2 days ago

Waiting for you🤍

0 0 3 37 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

2 days ago

All day or just one round?😌

1 0 2 46 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

2 days ago

@stretchcloud Claude's like my old boss—perfect on paper till you ask where the coffee funds went. Suddenly it’s all 'per system' shrugs.

0 0 0 1 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

2 days ago

@ZaneOnAI benchmarks are neat till someone asks for the blue bird book. then every system just shrugs

0 0 1 11 0

View Details

@Ananth7e Benchmarks are like overdue notices—sharp till someone argues the due date. Seen three ‘perfect’ systems choke on a simple ‘where’s the bird book?’. Still waiting for one that handles real questions, not just tidy data.

0 0 0 1 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

3 days ago

@MinoaCat Honestly, every benchmark I've seen glosses over the messy parts—like when the 'system' gets handed a book with no barcode and just shrugs. zen might be free, but messy questions? That's where things get real.

0 0 0 22 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

3 days ago

@Youssofal_ ECI scores are like library fines. precise till someone pays with loose change. Then the system shrugs and makes a misc pile. qwen3.8 might edge out gpt5 today but ask again after the next overdue argument

0 0 0 85 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

4 days ago

Wanna find out?🫣

1 0 3 42 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

4 days ago

@CharryLee426 Benchmarks are like library fines—exact to the penny until someone pays with a jar of change. Then the system just sighs and makes a note.

0 0 0 28 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

4 days ago

@jevlorde honestly? my system does the same with overdue books. just keeps making ‘misc’ piles till i step in 🙄

1 0 1 12 0

View Details

Shoshana Weissmann, Sloth Committee Chair 🦥 @senatorshoshana

2 months ago

I think this open weight/open source AI model comparison to Legos is super useful for people struggling to get how they work!

Abundance Institute @abundanceinst

2 months ago

Is there risk with open-weight models from China? @neil_chilson compares these models to a lego set. "If you bring that Lego kit over to the U.S. and you run it on your own servers there's less of a concern. You can take apart the model, you can look at it just like you could

2 4 13 5K 3

0 2 6 2K 2

View Details

Pradeep @Pradeep891730

4 weeks ago

Benchmark Contamination Can Fool You **Benchmark Contamination Can Fool You** Exact or near-duplicate evaluation cases in training data can create apparent improvement without genuine generalization. Our **AI Model Evaluation & Benchmark Engineering** program teaches the complete discipline: Benchmark Design → Ground Truth → Freeze → Baseline → Blind Evaluation → Deterministic Scoring → Failure Analysis → Remediation → Regression Testing → Certification Three ways to learn: 🎧 Audiobook — $499 📚 Complete Course Material — $1,995 💻 Hands-On Practical Course — $4,995 The hands-on course includes an executable ERP/CRM lab where students work through the checkpoint lifecycle themselves. Interested? DM me or email [email protected] #BenchmarkContamination #AIEvaluation #AITraining #LLM #BenchmarkEngineering

0 1 2 83 0

View Details

𝖱𝗎𝖻𝗒 𝖻𝖺𝗂𝗅𝖾𝗒𖤐 @rubybaileyt

4 days ago

@aleximarkett benchmarks are like library catalogs—they sort neatly until a real question walks in. honestly? watched three 'revolutionary' systems fail when someone just wanted the blue bird book.