marc

founder @ happyscribe

[marc @massens]

Spending a few weeks in SF makes the information gradient obvious:

labs → SF startups → X → Europe → traditional companies

Each step is months behind the one before it.

Sitting in SF building for a legacy industry is the easiest (and totally real!) edge right now.

[marc @massens]

Cool event for catalan founders currently in SF @Catalonia_TI

https://t.co/PecOofzflF

[marc @massens]

When transaction costs are nearly free, the boundary and shape of the firm will move. Can't wait to see the new kinds of companies this creates.

[marc @massens]

If markets are so efficient, why do companies exist? Why isn't everything just individuals transacting with individuals?

[marc @massens]

Perhaps meritocracy was never a philosophical problem. It was a technical one.

[vicc @viccpoes]

built my own version of after effects native to agents.

- the agent produces motion in fully editable layers and it can improve the tool itself by adding new features or controls on the fly.
- each project becomes its own version of the tool, with custom extensions, layers, or effects that the agent implements.
- native macOS app, so it can handle high resolutions and complex workflows with dozens of layers/effects.

planning to open-source, dm if you want to give it a try!

[marc @massens]

There's countless SDKs for humans to program agents

Where's the SDK for agents to program humans?

[marc @massens]

@danipolymath same feeling!

[marc @massens]

My agents now behave like my best hire:

→ full context, written down
→ 2–3 options
→ a clear recommendation
→ waiting for my call

I just open one page and make decisions.

[marc @massens]

@seeeeaaaannnnnn What’s the rise at the end?

[Sean Zhang @_seanzhang_]

We built the strongest attention-free base LLM, Brumby-14B-Base, for only $4,000.

This is just a base model, but for the first time in 2 years, I'm seeing light at the end of the tunnel for the field to switch to an alternative architecture to transformers.

[Manifest AI @manifest__ai]

Today we are releasing Brumby-14B-Base, the strongest attention-free base model around.

https://t.co/mclQPFdOGa

[marc @massens]

mindblown

https://t.co/cxhIFXS3aj

[marc @massens]

Really interesting concept https://t.co/zZxnUbbkd9

[marc @massens]

Just learned that we don't get 30% of ChatGPT traffic because it hallucinates our urls 🤡

[marc @massens]

#36 not bad! there's more bootstrapped companies than I expected

[Olivia Moore @omooretweets]

🚨 Introducing the AI Apps 50: Startup Edition

Ever wondered how startups are spending their money when it comes to AI?

Our team at @a16z worked with @mercury to crunch the numbers and rank the top applications by spend.

The list + what we learned from it👇

[Manifest AI @manifest__ai]

Today, we’re releasing Power Retention, a new architecture beyond Transformers.

It enables LLMs to handle millions of tokens efficiently, unlocking long-context applications that were too costly before.

https://t.co/Mdrgz3uBVX

[marc @massens]

i’m using @interaction for it because it has (1) memory and (2) push notifications.

i explore topics and ask it to create flashcards with what i want to remember directly on whatsapp. then every day 8.30 am it sends me a notification with the first flashcard.

as i practice and new questions arise i explore them in the middle of the session and create new flashcards.

the workflow does change a bit, instead of having lots of micro-flashcards on one topic like @michael_nielsen recommends, i use fewer and higher level flashcards, and i may do a 30s voice message to answer one, and then get feedback on details i got wrong or missed

it feels less like flaschards, and more like my mum questioning me before an exam when i was in school about a topic she understood well