davidbrianburch

Content marketer. I like hot dogs and freedom.

[vincent sunn chen @vincentsunnchen]

Senior SWE-Bench measures "tasteful solves": solutions are graded on not just correctness, but quality metrics (i.e. code quality, codebase practice alignment)

Tasks are based on real commits by maintainers with 100s of commits, in modern repos. They represent realistic "senior SWE jobs", e.g. "Design-and-Implement" a new feature, "Investigate-and-Fix" a multi-service bug

We introduce new reward mechanisms to enable more realistic tasks: a validation agent that adapts to different implementations and a taste judge that scores the "how"

Work led by @SnorkelAI w/ @karthik_r_n from the original SWE-bench team!

[Jason Lopatecki @jason_lopatecki]

Solid community 📚 paper reading coming up -- authors of the Self-Adapting LLMs (SEAL) paper, @adamzweiger and @jyo_pari of @MIT, answering your questions live https://t.co/iZ7U6LsbpV

[ADWEEK @Adweek]

VOICE | “As users grow accustomed to immediate AI answers, brand marketers face a parallel challenge of meeting these heightened expectations across every other touch point,” writes Brett Wilson.
https://t.co/5Dthjtl7rO

[davidbrianburch @davidbrianburch]

More in our latest blog: https://t.co/Q0pADxG7LO

[davidbrianburch @davidbrianburch]

Amid market volatility, how a company applies AI may just make all the difference -- particularly outside the walls of #GTC25. Here's a fun❓: which credit agency and online booking co are AI category leaders? You might be surprised by @swift_vc's AI Index picks. 🧵

[davidbrianburch @davidbrianburch]

In online travel services, $EXPE is a category leader in AI per @swift_vc's AI Index. Although it competes with AI innovators and pioneers like $BKNG and $TRIP that also have gen-AI use cases, Expedia's sizable investments in AI and hiring of related technical roles stand out.

[davidbrianburch @davidbrianburch]

Fun 🧵 on AI investments in public markets for those looking to get closer to the consumer https://t.co/4VoPq52WLT

[swiftventurecapital @swift_vc]

In an era of of open models and (eventually) declining inference costs, we’re entering a new era of AI application. For AI-minded investors in public markets, this is likely something of a reset. 🧵

[brett wilson @bjwilson34]

Maybe don't sell your $NVDA stock just yet. Our take on why DeepSeek isn't the existential thread portrayed in the media today. https://t.co/BzrVoKUUDf

[davidbrianburch @davidbrianburch]

Source: https://t.co/R161agM44B

[davidbrianburch @davidbrianburch]

Per an analysis by @swift_vc, the top five best-performing AI stocks among companies that lead the industry in hiring AI engineers and scientists + related roles are:
$SOUN $NVDA $UPST $IDCC and $META.

[davidbrianburch @davidbrianburch]

🔥🧵 on why companies that contribute to AI research and OSS outperform

[swiftventurecapital @swift_vc]

Do companies that contribute to AI research and open source tools and models outperform their peers in the public market? Here are a few reasons why it is a really powerful signal and contributes to our recently-launched AI Index's outperformance. 🧵👇

[brett wilson @bjwilson34]

Proud to debut the AI Index, a first-of-its-kind index of true AI companies in the public market. More on what we’re debuting; would love your thoughts! https://t.co/IHJfiJAxrf

[davidbrianburch @davidbrianburch]

🎆!

[VentureBeat @VentureBeat]

A new benchmark for AI investment: Swift Ventures unveils system to separate talk from action https://t.co/oWff3wxlJZ

[brett wilson @bjwilson34]

While some are saying the "scale is all you need" era might be coming to an end, we think the days of massive jumps in AI capabilities and productivity gains are just getting started. Here's why.
https://t.co/rfMN03AExe

[davidbrianburch @davidbrianburch]

Try it: https://t.co/VmAjkzOdOa

[davidbrianburch @davidbrianburch]

Most AI note takers are built to run great on Zoom, but the truth is many important meetings -- and huge sectors of the economy -- happen face-to-face. @LazyNotesapp by @stokaace is awesome for in-person meetings (or next to a laptop speaker) -- hit record on your phone and it transcribes and summarizes per your specific instructions, emailing you notes. I do a walkthrough here:

[davidbrianburch @davidbrianburch]

👀

[Arize AI @arizeai]

Almost half (45.9%) of AI engineers and other devs are using LLMs to code personal projects, and over one in ten use LLMs to automate busywork or create social media posts. 3.4% even admit to using gen-AI on dating apps!

[davidbrianburch @davidbrianburch]

27% of Fortune 500 companies cite AI regulation as a risk factor. Great piece by @bellelin_, with a minor assist from @arizeai. https://t.co/ZJsCK4P9lh