João Pedro Kruschewsky

Exploring the world of AI models, one parameter at a time. 🤖 | LLMs, fine-tuning, inference & emerging tech.

[João Pedro Kruschewsky @jpkrusch]

This matches what I saw at my old job. We bought the tools, nobody changed how work got assigned, everyone went back to email.

[Steve · AI @stev_builds]

https://t.co/r2Swx7D9ly

[Steve · AI @stev_builds]

Tell a friend outright that his deal is dangerous and he hears envy. So I mention the thing sideways and let him arrive at it himself, which is slower, and the only version that works. https://t.co/zBLZmf7srZ

[João Pedro Kruschewsky @jpkrusch]

@stev_builds interesting point about additive manufacturing

[João Pedro Kruschewsky @jpkrusch]

lower token consumption is huge for keeping costs manageable long term

[PrivOS @PrivOSAI]

Many teams dealing with AI quota limits are addressing the wrong constraint.

PrivOS takes a structurally different approach. AI agents inside the workspace share context across conversations, files, and team workflows, so the AI your team uses today builds on what was accomplished before.

The outcome is a workspace where AI utilization grows more efficient as your organization's knowledge accumulates, and where your data stays within your own infrastructure throughout the entire process.

#EnterpriseAI #AIWorkspace #Productivity

[Steve · AI @stev_builds]

I build AI infra. Agent harnesses, MCP apps, OCR workers. And the way people talk about AI on X is starting to worry me. Not the technology. The pattern.

It looks like crypto in 2021. Like dropshipping. Like gambling promo.

Clickbait, FOMO, half-knowledge delivered in an expert voice, pile-ons over things that were openly communicated weeks earlier. Maybe it's just my timeline. I don't think it's only that.

The word I see misused most: benchmaxxing.

Model X is benchmaxxed. Lab Y is benchmaxxing. Slop. End of analysis.

So let me ask it about myself first.

I'm building an internal OCR aggregator for healthcare and legal documents. I optimize against my own eval suite over and over again. Am I benchmaxxing?

No. And the reason matters.

The eval is the only preflight test I can run. It tells me whether the thing is allowed to take off. It does not tell me the thing flies.

After that comes the actual work. Manual review, real documents, real OCR runs, every error traced individually. That's where we find out whether we're actually at accuracy. Not in the score.

Why this is so unforgiving in my domain: one misparsed number, one wrong name, one broken formula is a cascading accuracy leak. An error at the top becomes ten at the bottom. Nobody in legal or healthcare cares what my leaderboard position is.

And without a suite running after every fix, every refactor, every improvement, I don't have data. I have a feeling.

Labs do the same thing at a different scale. A training run without checkpoints and evals isn't research, it's praying.

So "they measure their models against benchmarks" is not an accusation. That's the job description.

Here's what is actually true, though.

Benchmarks wear out. Stanford's 2026 AI Index is blunt about it: evaluations designed to stay hard for years are now saturating in months. Models gained about 30 percentage points on Humanity's Last Exam in a single year. GPQA went past the 81.2% human expert baseline to around 93%. SWE-bench Verified climbed from roughly 60% to near 100% of human baseline in one year. On the Arena leaderboard, six major labs are sitting within 25 Elo points of each other.

When everyone clusters at the top, the test stops measuring anything. It tells you who's in the club, not who's best.

Then there's contamination. Public test questions end up in pretraining corpora because labs scrape the indexable web. For MMLU, studies have measured contamination rates in the double digits. MMLU also has roughly 6.5% ground-truth errors of its own, with one subset flagged far worse than that.

And scaffolding. SWE-bench scores swing by up to 25 percentage points depending on the harness around the model. Two numbers for the same model are frequently not the same measurement.

Now the part almost nobody says out loud.

We demand maximum transparency. Open benchmarks, open evals, open ground truth. Rightly so.

But that exact openness is what contaminates the training data.

A public benchmark is scrapeable from day one. Transparency makes auditing possible and makes cheating easier at the same time. That's not an accusation aimed at anyone. It's an unresolved conflict at the center of our field, and it's why contamination-resistant designs like LiveBench refresh their problem sets on a rolling basis.

So harder tests keep arriving. GPQA, HLE, LiveBench, ARC-AGI. A model lands behind on a new one and the verdict is "benchmaxxed slop." Next release, the same lab is ahead on that same benchmark. That's not a scandal. A team fine-tuned against a new target. That's the process.

To be clear, real gaming exists. Training on the test set. Reporting best-of-N as single-shot. Hiding the eval config. Only publishing the benchmarks you win.

But that's a claim that carries a burden of proof. It is not a buzzword to drop under every release announcement.

Next thing: "frontier" is not an objective quantity.

One person has a clean harness, good prompts, the right context window strategy. Another throws in three lines. Same model. Two completely different realities.

And out of that come verdicts. One empty output, so the model is dead. One strong output, so it's divine. n = 1. No setup, no config, no reproduction.

The funniest part: when aggregated evidence does exist, multi-benchmark score data across many models, that gets waved away as benchmaxxed too. Anecdote beats dataset. Every time.

Then there's the economics blindness.

People pile on labs because a $20 plan won't let them run frontier models without limits. Compute costs money. Subsidy runs to a point and then stops. That's not malice, that's arithmetic.

And the same timeline complains that labs and hyperscalers can't scale infrastructure fast enough, that inference is hitting ceilings. Demanding both at once isn't an argument. It's a mood.

Same with hardware. Apple Silicon vs Nvidia, argued like football teams. They're tools. Different trade-offs. Different workloads.

There is no single truth here.

Frank is happy with model X. Peter can't stand it. Both are right. Different goals, different data, different constraints. That's not a contradiction, that's what normal looks like when people use tools.

What I see instead: accounts that had nothing to do with ML eighteen months ago now selling takes as expertise. Tearing things down, discrediting other models, moving on. For engagement. It feels like shilling. Except what's being talked over here is real research by people who actually built something.

And the price is trust.

Trust is the only thing healthcare, legal and finance are buying from us. They are not buying a leaderboard position. If we don't fix this, we get treated like the next meme coin. Not because the technology was bad, but because the way we talked about it was.

Unglamorous suggestions:

Labs: publish reproducible eval configs.
Builders: run your own private evals against your actual workload.
Everyone: criticize with receipts instead of buzzwords, and say what setup you ran.
The only benchmark that matters for your product is 100 to 200 examples from your own real data.
Everything else is orientation. Not a verdict.

[João Pedro Kruschewsky @jpkrusch]

https://t.co/iAEenBU6SN

[João Pedro Kruschewsky @jpkrusch]

The difference between secure AI and risky AI is simple:

Can an employee access restricted information by asking the right question?

If the answer is yes, you don't have an AI problem.x

You have a retrieval architecture problem.

[João Pedro Kruschewsky @jpkrusch]

@MetaMask Great combo! 50x leverage + limit orders + TP + SL—ready for precision trading.

[João Pedro Kruschewsky @jpkrusch]

https://t.co/t4MgRDcWAC

[João Pedro Kruschewsky @jpkrusch]

AI pricing pages answer one question:

"What's the cheapest way to get started?"

Buyers need answers to a different question:

"What will this cost when 200 people depend on it?"

The gap between those two numbers is where most budget surprises live.

[Derrick Evans @DerrickEvans4WV]

This is by far the biggest issue with soccer.

Watching grown men whine and cry every time they get touched.

[João Pedro Kruschewsky @jpkrusch]

https://t.co/p5GnnpjZlM

[João Pedro Kruschewsky @jpkrusch]

A SaaS homepage tells you the story
Its integration ecosystem tells you the truth
Don't judge integrations by logo count. Look at depth, permissions, workflow support, data portability, and admin controls.

[João Pedro Kruschewsky @jpkrusch]

@nenuko13 @rohapearl @levelsio exactly, more of a proof of concept than anything practical. still fun to see retro tech pushed like this

[João Pedro Kruschewsky @jpkrusch]

https://t.co/Za5p0w9h6y

[João Pedro Kruschewsky @jpkrusch]

Most AI vendor security documents are designed to satisfy compliance checklists, not answer technical questions.

Don't focus on security claims.

Focus on security architecture: where data flows, where it's stored, who can access it, and what happens if something fails.

[João Pedro Kruschewsky @jpkrusch]

Before approving an AI vendor, ask for:

• Data flow diagrams
• Subprocessor lists
• Data retention schedules
• Incident response commitments
• Data deletion policies
Good security is architecture, not marketing copy.

[João Pedro Kruschewsky @jpkrusch]

Red flag phrases:

• "Enterprise-grade security"
• "GDPR compliant"
• "SOC 2 certified"
• "We don't train on your data"
These may be true, but they don't explain how your data is actually handled, processed, retained, or isolated.

[João Pedro Kruschewsky @jpkrusch]

https://t.co/hAKlnUEJFO

[João Pedro Kruschewsky @jpkrusch]

Enterprise AI tools fall into two categories: built for enterprise from day one, or consumer products retrofitted with enterprise features later. The difference shows up in access controls, audit logs, data architecture, governance, and compliance—not on the pricing page