Kate | Neuromancer

GTA veteran. FIFA grinder. 2K survivor.

tech girly, i explain what's happening in AI, startups, and gaming and tell you exactly what i think about it

[Kate | Neuromancer @Neural_Reviewer]

when every ai tool starts having issues at the same time 😭

openai, claude, grok, cursor…

guess we're all touching grass today

[Kate | Neuromancer @Neural_Reviewer]

@DrKoWN_ @TheMW3Goat @wccftech 😅

[Kate | Neuromancer @Neural_Reviewer]

@NoahStephanoss

[Kate | Neuromancer @Neural_Reviewer]

saw a founder tweet this morning: "a $1M-$5M acquisition can completely change your life."

and honestly? they're right.

startup twitter broke everyone's brain on this. you see a $50M acquisition and assume everyone got rich.

then you realize the founder owns 10% after multiple rounds of dilution. that's $5M pre-tax. maybe less after liquidation preferences.

meanwhile some bootstrapped founder sells for $3M and owns 90% of the company. that's $2.7M. cash. no investors to pay back. no preferred shares to satisfy.

do the math.

the headline number doesn't tell you much. ownership percentage tells you everything.

keep your ownership. build something profitable. don't feel like you need a massive exit for it to be a win.

a few million in the bank is a pretty good outcome.

[Kate | Neuromancer @Neural_Reviewer]

@cb_doge private jets getting reliable satellite internet while flying around the world feels very on-brand for 2026

[Kate | Neuromancer @Neural_Reviewer]

@verge robot lawnmowers finally learning that the edges exist

built-in edgers sounds like one of those tiny upgrades that makes you wonder why they weren't standard already

less lawn cleanup after the robot does its job = i'm listening.

[Kate | Neuromancer @Neural_Reviewer]

@DaRealCyberLeek

[Kate | Neuromancer @Neural_Reviewer]

@code pull requests + agents + related chats i one release?

vs code is getting very serious about how we actually work with code now

and a whole story of vs code premiere on september 4 👀

[Kate | Neuromancer @Neural_Reviewer]

@cursor_ai gemini 38 flash landing in cursor already 👀

this is the kind of rollout i like — new model, straight into a tool people are already using

now i wanna see how it actually feels in a real coding session.

[Kate | Neuromancer @Neural_Reviewer]

@paulg this is probably the best startup advice you can get 😭

sometimes the fastest way to grow is to stop building and spend all your time talking to customers

22 customers > another six months of features

[Kate | Neuromancer @Neural_Reviewer]

gamescom awards are out. cd projekt red and sega took home 3 trophies each.

cdpr won for the witcher 3 remastered announcement, songs of the past dlc trailer, and best booth. sega won best audio for alien: isolation 2, best pc game for total war: warhammer 40000, and best lineup.

capcom's onimusha: sword of destiny won best gameplay and best playstation game. metro 2039 won best visual.

red desert got nominated in two categories and walked away with nothing. rough.

the witcher 3 remastered drops september 29 as a free upgrade for existing owners. that's next month.

[Kate | Neuromancer @Neural_Reviewer]

fable 5.1 beating fable 5 across the board on document parsing is actually interesting

tables + formatting + visual grounding are the kinds of improvements i'd care about in practice

the $0.15/page price tag though… yeah, that's a different conversation 😭

[Jerry Liu @jerryjliu0]

Fable 5.1 didn't come with document OCR benchmarks, so we benchmarked it on ParseBench: a comprehensive parsing benchmark that's comprised of 2k+ real-world documents across finance, legal, insurance, and more.

As a positive note, it is one of the first frontier models in recent memory where document OCR performance actually improved between model generations!
✅ It beats Fable 5 across the board on all of our metrics, from tables/content faithfulness/formatting/charts/visual grounding
✅ It is one of the best models out there for parsing tables

This is great, because while frontier models have rapidly advanced on reasoning benchmarks, they have not advanced on visual understanding. Prior to this the latest models (Opus 5, 5.6-Sol, 3.7 Flash) have stagnated on visual understanding tasks.

Obviously you shouldn't use Fable 5.1 as a document OCR tool on its own; besides missing needed metadata, it also costs 15c a page - which is 10-15x more expensive than our default agentic mode on LlamaParse (which performs similarly).

If you're interested in tracking the 90+ solutions that we've benchmarked on doc OCR, check out ParseBench: https://t.co/PWczfhosZp

If you're looking for a dedicated, SOTA OCR tool to power your doc processing workloads, check out LlamaParse: https://t.co/XYZmx5T7JA

[Kate | Neuromancer @Neural_Reviewer]

@testingcatalog

[Kate | Neuromancer @Neural_Reviewer]

@sundarpichai 86.2% on CyberGym + 47.2% on CWE-Bench while keeping Flash-level speed and pricing?

okay, cybersecurity is getting very serious 👀

70%+ vulnerability discovery across 20 languages is probably the part i'd be watching closest.

[NotPalo @NotPaloIntel]

Rockstar Games finalmente ha actualizado la fecha que aparece en los Tráilers 1 y 2 de GTA VI.

Ambos ya muestran el 19 de noviembre de 2026, la nueva fecha de lanzamiento del juego.

[Kate | Neuromancer @Neural_Reviewer]

@GTA6Tracker not that i'm counting.

[Kate | Neuromancer @Neural_Reviewer]

2.6x more correct Chrome vulnerability patches than much lrger models?

okay, that's thekind of benchmark result that gets my attention 👀

the lower cost + higher recall from wiz is interesting too. i'd want to see how this holds up outside internal testing.

[News from Google @NewsFromGoogle]

How does Gemini 3.8 Flash Cyber hold up against real-world threats?

We've been stress-testing it internally at @Google, and here's what we saw:

🌐 @GoogleChrome: The Chrome Security team found that 3.8 Flash Cyber produced 2.6x more correct patches to vulnerabilities in Chrome than the best commercial models that are much larger.

☁️ @GoogleCloud: The Google Cloud Vulnerability Research team detected a critical foundational vulnerability in less than 2 hours — research and discovery that typically takes months.

🛡️ @wiz_io: Achieved +7.5-9.7% higher recall on their internal penetration testing benchmark for a 2.3-5.2x lower cost compared to other leading frontier models.

[Kate | Neuromancer @Neural_Reviewer]

78 days until gta vi.

rockstar just dropped a nearly 27-minute extended look and somehow the internet still wants to analyze every frame like it's the zapruder film.

november 19 cannot come fast enough.

78 days. not that i'm counting.

[Kate | Neuromancer @Neural_Reviewer]

helsinki-based nimble fox is launching an ai tool for unity developers on september 7 .

here's the difference from other ai coding tools: it doesn't generate a full game in one pass. it lets you build feature by feature. implement gameplay mechanics, ui flows, one at a time. reads the context of your existing project and makes changes through natural language .

the founder jonne myöhänen started it because he was frustrated with how long prototyping took. turned it into a company, raised funding, got a grant from business finland .

they're also hosting a 48-hour game jam from september 4-6. participants get free access to the tool .

the approach is interesting. instead of "write my whole game for me" it's "help me build this specific feature." that's a more realistic use case for actual developers.

https://t.co/7gw0qB93sH