seb @sebcrossa

cofounder @llmstats (yc s25). prev. @microHQ sebastiancrossa.com nyc Joined February 2015

Tweets

678
Followers

634
Following

2K
Likes

2K

seb @sebcrossa

5 days ago

@ay_ushr thanks, stealing this

0 0 1 65 0

View Details

the problems are everywhere: data contamination (training on test sets), selective reporting (only publishing the good scores), prompt tuning that users can't replicate, cherry-picked metrics, custom internal benchmarks with zero transparency.

0 0 0 8 0

View Details

seb @sebcrossa

6 days ago

kimi k2 reported 50% on humanity's last exam. independent researchers tested it and got 29.4%. that's a 20 point gap on the same benchmark, same model.

1 0 0 26 0

View Details

seb @sebcrossa

a week ago

@n0w00j proud of this achievement

0 0 4 467 0

View Details

seb @sebcrossa

a week ago

@artists_voyage slop

0 0 1 69 0

View Details

seb @sebcrossa

a week ago

started using poke two months ago and haven't gone a day without texting it. it's crazy good and 100% think people should try this before trying to set up openclaw / hermes

Poke @interaction

2 weeks ago

Say hi to the new Poke! 🌴 Now officially approved by Apple to text on Apple Messages. As the first and only AI agent. Chat now: Poke.com

435 248 5K 4.9M 3K

5 1 27 4K 2

View Details

seb @sebcrossa

a week ago

@aryaan_smm check again

0 0 2 68 0

View Details

seb @sebcrossa

a week ago

5b tokens cursor.com/@sebcrossa

eric zakariasson @ericzakariasson

2 weeks ago

introducing cursor profiles! go claim your handle at cursor.com/profile

355 96 2K 704K 724

0 0 0 86 1

View Details

seb @sebcrossa

2 weeks ago

@benkimbuilds 100% true, have seen this firsthand

1 0 1 539 0

View Details

seb @sebcrossa

2 weeks ago

we started using them a few days before a big weekend outage. big fan of what @shcallaway and the team are building! in an era of slop, it's relieving to see a team with such high standards for craft and such crazy product velocity.

Y Combinator @ycombinator

2 weeks ago

Sazabi is a next-generation observability platform designed for fast-moving, AI-native engineering teams. If you're using Datadog, Sentry, Grafana, or Axiom today, you could be moving ten times faster with @sazabi. Congrats on the launch, @shcallaway! ycombinator.com/launches/QdT-s…

32 22 227 66K 81

1 1 7 1K 0

View Details

Y Combinator @ycombinator

2 weeks ago

32 22 227 66K 81

View Details

seb @sebcrossa

2 weeks ago

one of the first things you learn when starting a company is that it's 100x harder than you realized. but the greatest side effect is that you start noticing the hard work that was put into everything around you. from the businesses, the ads, the buildings. everything you see was built by people who poured their blood, sweat and tears into making things possible.

0 0 1 54 0

View Details

LLM Stats @LlmStats

3 weeks ago

PROGRESSION: Claude Opus 4.8 (@AnthropicAI) sets a new high on LLM Stats Index after 3 years of progress. > 68 on LLM Stats Index > Frontier gained 85 index points over the 3.3-year window This shifts the predictions towards a more optimistic scenario of complete benchmarking saturation.

2 2 13 903 0

View Details

LLM Stats @LlmStats

3 weeks ago

NEW #1: Claude Opus 4.8 (@AnthropicAI) takes the top spot on LLM Stats Index. > 68 on LLM Stats Index > +5 over previous SOTA (GPT-5.5)

0 3 5 425 0

View Details

LLM Stats @LlmStats

4 weeks ago

NEW: Gemini 3.5 Flash (@GoogleDeepMind) lands at #5 on LLM Stats Index. >4x faster than other frontier models >$1.50 / $9 per 1M tokens It's significantly faster than other frontier models and the quality has increased significantly. It's also the best tool calling model we've tested lately.

3 2 5 819 0

View Details

seb @sebcrossa

a month ago

@mil000 wait til you hear about pivots

0 0 0 28 0

View Details

LLM Stats @LlmStats

a month ago

@polynoamial Hey Noam! We're in the process of building new hard benchmarks focused on coding/long-ctx. We're a team of 2, fully focused on AI capability tracking since last year. Is it possible we could talk with someone from your team to align on what is maximally useful to measure?

0 1 8 624 0

View Details

Manufact @manufact

a month ago

🚀 LAUNCH WEEK KICKOFFS TODAY! The fastest way to ship an MCP app: Scaffold with mcp-use. Deploy with Manufact. Ship to Claude and ChatGPT. 🧵Here's the full path: 1/5

10 9 33 3K 7

View Details

LLM Stats @LlmStats

a month ago

NEW: GPT-5.5 Instant is now available on LLM Stats. Try it now for free in our agent and code playgrounds.

1 1 4 326 0

View Details

LLM Stats @LlmStats

a month ago

Today we're introducing the LLM Stats Index. For 3.2 years, we've tracked every frontier model release. The Index aggregates 200+ benchmark results into a single TrueSkill rating per model, spanning law, healthcare, coding, tool calling, vision, and reasoning. Across every category and every modality, the leading model on the Pareto Frontier is GPT-5.5 (@OpenAI). On our trajectories, human-knowledge benchmarks saturate by mid-2027. Capability has been the primary axis. The field is converging on it. Two more are opening. The first is efficiency: total task cost is the cleanest proxy we have for intelligence/watt. The second is throughput: inference speed becomes the productivity ceiling once models are cheap and good enough. We're building the next generation of long-horizon coding, tool use, and long context benchmarks. If you're working on long-horizon evaluation in real domains, we'd like to chat.