mblevin

[mblevin @mblevin]

@Hadley Requires being a builder though. Impossible to be an expert without actually building agents / agentic systems and hitting all the edge cases yourself. Likewise the jagged frontier means that knowledge is out of date with every release and you must be in the trenches again.

[mblevin @mblevin]

@ajhodls Read this thing from another investor who prefers the opposite: https://t.co/5DbIoHiWBg

[first check $500k-1M pre-seed @ajhodls]

one of the biggest alpha opportunities in early-stage VC is low charisma founders who don't pitch well, don't tweet, don't podcast, but just execute day in day out.

[mblevin @mblevin]

@ilyasu Incredible

[mblevin @mblevin]

@signulll Why would you think this not just an in-house labeling platform with some marketing spin?

[mblevin @mblevin]

@OfficialLoganK After you caused a panic attack among many startups and hours of wasted root cause investigation into what appeared to be a major key leakage issue on a Sunday

[mblevin @mblevin]

@simonw @ileppane Half of this is downstream from being too cute by half with model naming. Some of which I suspect was coming from the CEO office with bad judgment around consumer understanding.

[mblevin @mblevin]

@BucknSF This implies Figma Make is a zero, which I think is tough bet to make.

[mblevin @mblevin]

@brackin @BucknSF I honestly wonder if he misspoke. It’s so wildly opposite of the currently lived reality of AI native builders, buyers etc.

[mblevin @mblevin]

@yoheinakajima I’m skeptical. “store useful memories, and at query time grab relevant ones and stick into the context window” has no moat unless none of it is accessible to the user, which seems unlikely. Also easily replicated by a de novo setup agent that has access to email, calendar, etc

[mblevin @mblevin]

@modestproposal1 Unsurprising in the least to people who actually build products with LLMs. I’d be shocked if the distribution was anything but.

[mblevin @mblevin]

@teddypowday [response based on even less information]

[mblevin @mblevin]

@JStein_WaPo @JessicaBRiedl Who says anything they've identified is waste, fraud or abuse? Come on.

[mblevin @mblevin]

@aashaysanghvi_ Do any SDRs get comped solely on closed won revenue?

[mblevin @mblevin]

@quantian1 There’s no way to accumulate that kind of capital over time as a trader with that kind of doofus logic. Yes clearly lying about backers and/or personal capital.

[mblevin @mblevin]

@natfriedman If the input device switches (like your AirPods disconnect for a second) Granola stops recording, there is no way to restart, and you’ve lost the transcript and recording, which sucks big time.

[mblevin @mblevin]

@natfriedman Great UX model; super buggy and transcripts and action items way behind Circleback. If they solve those things though they are the clear category leader.

[mblevin @mblevin]

@conorsen I don't think that's true across the board at all. Many folks correctly identified petrodollars blowing a massive bubble in stocks and housing combined with a variety of poor financial practices - "big short" style synthetic CDOs, no-doc loans, etc. Unrelated to long-term demand.

[mblevin @mblevin]

@aashaysanghvi_ Can’t be universally applicable right. Seems like it only applies in substituting for high cost premium white collar labor (lawyers, accounting, consulting, proposal / RFP response generation , etc).

[mblevin @mblevin]

@paulg But he’s actually outed himself repeatedly over years. More important question - why not been obvious to otherwise very smart, evidence-based people?Ex: https://t.co/C3BpRHwYUI.

[mblevin @mblevin]

@finbarrtimbers 1. Huge platform shift - have to be in it somehow. Foundational models are like building a browser or smartphone.
2. Commodifying potential competition that could emerge
3. Clear utility across different parts of Meta offerings (image / text gen) - bad biz to pay rent to OAI.