kyle | kbw.Ξth

What’s Next 🤔 Co-founder @humnPassport, Co-founder @Gitcoin, Co-founder @ColoradoProduct. Alum of @Twitter. Just here to buidl

[Sriram Krishnan @sriramk]

It is clear open source models and harnesses are having a moment. There's a few factors at work

1/ It is now obvious that you can catch up to near-SOTA performance and do so with a clear training lineage. See:@thinkymachines Inkling launch today.

2/ There are several well-funded, talented teams building open weight models now in the US and abroad. Along with the explosing of other near SOTA models (Grok/Cursor, Muse Spark), it is clear we are going to have a diverse ecosystem of models atleast on coding and agentic use.

3/ Organizations are increasingly looking for control over how their data is used and are willing to trade off some access to frontier level tokens for this control. Organizations and countries are increasingly nervous about the frontier labs potentially competing with them down the road and don't want their data to enable a future competitor.

4/ Open source is a slider: you could bring your own open harness, your evals, your business context and are free to pick and choose your model of choice.

5/ Companies have now actively shifted from "how do we get our people to use tokens" to being uncomfortable with their token cost ballooning without a clear line to revenue.

6/ Geo-politically, countries will be weighing open weight models as a way to get frontier-level tokens inside controlled environments that may not be otherwise possible.

All of this leads to more choice for all of us !

[kyle | kbw.Ξth @kbw]

Fable + Imagination + Friday Vibes 🎉🥳

[kyle | kbw.Ξth @kbw]

https://t.co/Zgs8Hdcp8t

[kyle | kbw.Ξth @kbw]

Been playing with Claude Tag for a handful of hours across three large organizations (100-2k users). Some initial thoughts:

The new Claude AI Tag is interesting… I am not sold on it though. It starts to take the concept of a Git based “brain” and turn it into a Claude based “brain.” Hellloooooo lock-in concerns!

The permissions, and wiring make it powerful, but it abstracts all of the memory, knowledge and skills that the Git based company brain affords.

If you are rolling out Claude to novice users who live in Slack, it can be valuable… BUT, if you can skill up your team and get them into a department level brain (Git based), then Claude Tag is not the way to go.

Regardless, its fairly straight forward to configure, but watch out because if you give Claude access to things, fine grained auth scopes don't exist for everything yet (ie, give it access to see a repo, it can now commit and merge changes unless your repo has stop gaps in place).

Oh, and the pricing is steep on the per model token model 😬

[Claude @claudeai]

Introducing Claude Tag, a new way for teams to work with Claude.

In Slack, Claude joins as a team member with access to the channels and tools you choose. Tag Claude in and delegate tasks to it while you focus on other work.

[kyle | kbw.Ξth @kbw]

Seems like memory systems and management is going to be the next ai-race.

Open system that’s portable between LLMs will be powerful. And I don’t think obsidian is it.

[kyle | kbw.Ξth @kbw]

for those building "company brains" what harness are you choosing for non-technical users?

Hermes with whatsapp/Slack integration is the path we are testing and I wonder what else is out there.

[kyle | kbw.Ξth @kbw]

The pace in which openAI and anthropic can move upstream is just wild.

[kyle | kbw.Ξth @kbw]

Is your AI Consultant like Nick? AI adoption is a learning process... You don't need a Nick.
https://t.co/u5sWRdsQvi

[kyle | kbw.Ξth @kbw]

The build log is accelerating as the number of items being delivered also grows. it feels good to be moving faster with more complex items

https://t.co/capxJwEPKP

[kyle | kbw.Ξth @kbw]

Let's use AI to generate stuff that doesn't look like AI generated it 🙃

[kyle | kbw.Ξth @kbw]

This is spot on... the era of mass adoption will come from simplification/normalization of the power that the coding focused tools bring.

[Peter Yang @petergyang]

https://t.co/o30TvRVbAz

[kyle | kbw.Ξth @kbw]

Love this take. The optimism is palpable... though may be laced with a bit of hopium.

[David George @DavidGeorge83]

https://t.co/QXVYrAv6H4

[kyle | kbw.Ξth @kbw]

This is really solid. Iv'e found we need a couple more things: Crystallization Ladder, Tiered Rigor/confidence, content-type trust signals... some of that can be woven into the memory traces, but are functionally different.

[Ashwin Gopinath @ashwingop]

https://t.co/7mNAkvprLH

[kyle | kbw.Ξth @kbw]

@DisruptionJoe MCPs do error on token heavy as folks developing them feel more context is better. sometimes it is, sometimes it isnt. ¯\_(ツ)_/¯

[kyle | kbw.Ξth @kbw]

I am starting to move away from a CLI... Codex and Claude Code in the mac desktop app are that good.

what am I missing? Is a CLI still valuable?

[kyle | kbw.Ξth @kbw]

@DisruptionJoe interesting, and is that a CLI specific thing... or a "use the API directly" thing?

[kyle | kbw.Ξth @kbw]

a PM asked me - "what's it like building agentic systems/products?"

I responded roughly - "remember how exhaustive and tedious writing some acceptance criteria was for junior devs? Now 10x that misery when you are writing and reviewing evals."

[kyle | kbw.Ξth @kbw]

@zherring the interface is just cleaner. I can see what files are in context, the branch i'm on, still have a terminal window, and quick access to artifacts I am working with.

It gives me loads of context, to then feed more context into my prompts. planning also seems better with the flo

[kyle | kbw.Ξth @kbw]

We are actually on the precipice on Web 3.0. It’s an agententic web.

Immutable, decentralized networks and information are not things the masses really care about (still).

[kyle | kbw.Ξth @kbw]

This about this opp for the Russel 2000... Don't have the tools to do this, but have the money and (hopefully) the drive to accomplish.

[Geoff Charles @geoffintech]

https://t.co/CaKyQS1Fl4