Walter Reade

8.2K posts

Walter Reade

Walter Reade

@WalterReade

AI at Google/Kaggle (inversion). PhD chemical engineer. Too many hobbies.

Neenah, Wisconsin 54956 Katılım Aralık 2007
2K Takip Edilen9.2K Takipçiler
Sabitlenmiş Tweet
Walter Reade
Walter Reade@WalterReade·
Is your child texting about @kaggle ?
Walter Reade tweet media
English
9
205
702
0
Walter Reade retweetledi
Joel Grus 🤠
Joel Grus 🤠@joelgrus·
"if you never hit your claude fable limit, you're spending too much time in airports"
Scenic Oaks, TX 🇺🇸 English
0
2
9
667
Walter Reade
Walter Reade@WalterReade·
@JFPuget We've always appreciated your blunt but always helpful feedback at Kaggle.
Neenah, WI 🇺🇸 English
1
0
2
319
Walter Reade retweetledi
Simon Jegou
Simon Jegou@simon_jegou·
I am finally a @kaggle Grandmaster after earning my first solo gold medal in the Orbit Wars competition! 🤩 Two years ago, I joined the KGMoN team at @nvidia as the first non-Grandmaster, so I guess my impostor syndrome has officially expired today 😅 (1/3)
English
12
4
56
7.3K
Mormonr
Mormonr@themormonr·
Watch our short film about the Cane Creek Massacre, when a mob attacked a worship service of the Church of Jesus Christ of Latter-day Saints in Cane Creek, Tennessee in 1884. Link in the replies.
Mormonr tweet media
English
5
8
64
5.7K
Dr Manhattva
Dr Manhattva@Manhattva·
I hope that after this trial, That all of Candace Owens’ followers realize that she is nuttier than squirrel shit.
English
273
168
3.7K
153.2K
Walter Reade
Walter Reade@WalterReade·
A year ago, when we were still coding by hand, it was always tempting to work on projects that were easy but less important. There's really no excuse not to be working on the important things now.
English
4
2
27
1.5K
Walter Reade
Walter Reade@WalterReade·
@svpino The challenge to navigate is that some engineers*should* be given a larger token budget that others.
English
0
0
0
143
Santiago
Santiago@svpino·
I've already talked to two companies that limit engineers to $100 in weekly tokens. That's around $20,000 in extra costs per person per year. After you run out, you gotta write the code yourself, like a caveman. I suspect most companies will end up here. There's no way they can offer an open buffet of tokens without going bankrupt.
English
173
21
478
78.1K
Toven
Toven@pingToven·
a picture is worth a thousand words
Toven tweet media
English
2
0
10
489
Connor Boyack 📚
Connor Boyack 📚@cboyack·
The HOA demanded putting up a fence to block the trash cans from view, so…
Connor Boyack 📚 tweet media
English
78
95
3.6K
48.7K
Walter Reade
Walter Reade@WalterReade·
It's amazing how quickly we can no longer live with the previous model we couldn't live without.
English
2
0
7
761
Bojan Tunguz
Bojan Tunguz@tunguz·
@WalterReade Fable is now running circles around the problem that's been flying in place for a week with Opus 4.8 and GPT 5.5.
English
2
0
7
655
Bojan Tunguz
Bojan Tunguz@tunguz·
We all get seven days to escape permanent underclass.
English
31
8
317
24.8K
Walter Reade
Walter Reade@WalterReade·
@alexgshaw TBF, that's how I would have done it. 🤷
Neenah, WI 🇺🇸 English
1
0
1
207
Alex Shaw
Alex Shaw@alexgshaw·
I asked an agent to display some info on Harbor Hub. It decided the best approach was to fetch the entire postgres table and filter and aggregate client side. We need to evaluate agents' abilities to apply best practices without being told. Excited for Senior SWE-Bench!
Henry Kiss Ehrenberg@henryehrenberg

We expect agents to act like senior engineers, but most benchmarks still evaluate them like interns. Excited to introduce Senior SWE-Bench, an open-source and @harborframework-native benchmark that assesses agents as senior engineers on long-horizon tasks with realistically under-specified instructions. We expect agents to build real features going on just a quick Slack message, nothing like the super technical instructions most benchmarks provide. Senior SWE-Bench fixes that. Claude Opus 4.8 is the current leader at 24% high quality solves, but it took 117K tokens on average to get there. Claude Sonnet 5 looked like it was going to swoop in for the top spot, but we found it cheated on 26% of trials.

English
3
7
46
4.1K
Walter Reade retweetledi
María Cruz
María Cruz@marianarra_·
Last week, @kaggle launched this initiative to promote local hackathon organizing to build with Gemma 4. If you host a hackathon for your local community, Kaggle will cover the prize pool funds for your hackathon winners 🤑 Apply by July 31st! 👉🏼 goo.gle/build-with-gem…
Google for Developers@googledevs

Ready to bring the power of open models to your community? The @GoogleGemma team is sponsoring 1-day hackathons on @Kaggle! Whether you want to explore how Gemma 4 can be used to build lightweight tools or drive AI innovation in a specific field, now's your chance to lead the charge. 👉 Apply to host a hackathon for your community today: goo.gle/build-with-gem…

English
0
2
6
585
Walter Reade
Walter Reade@WalterReade·
When I was a ChemE student at BYU, it wasn't uncommon to have closed book, timed, take-home exams. It's my understanding they've also had to change that practice due to AI. theatlantic.com/ideas/2026/05/…
English
1
0
1
365
Walter Reade
Walter Reade@WalterReade·
@joelgrus "red and yellow, kill a fellow; red and black, friend of jack"
English
0
0
0
117
Joel Grus 🤠
Joel Grus 🤠@joelgrus·
chatgpt earning its keep this week
Joel Grus 🤠 tweet media
English
1
0
1
509
Walter Reade retweetledi
meg.ai 🇨🇦
meg.ai 🇨🇦@MeganRisdal·
Really enjoyed this convo with @WalterReade & @alexgshaw! We talked about how Alex's work with Kaggle on rigorous evaluation for the $1M KPrize (& lessons learned) inspired Terminal-Bench inspired @harborframework.
Kaggle@kaggle

In the first episode of Kaggle Conversations, @MeganRisdal and @WalterReade from the Kaggle team talk with @alexgshaw, creator of @harborframework, about how the AI Agent industry is moving towards autonomy, why software engineering benchmarks are prone to contamination, and what it took to build a secure evaluation framework for the $1M Konwinski Prize. Watch a clip from the episode 👇

English
1
2
12
1.6K