Lotto

mlai consulting side gig / building localmaxxing

[Lotto @LottoLabs]

https://t.co/tawWejKZC2

[Lotto @LottoLabs]

@usr_bin_roygbiv Jesus this looks promising

[Lotto @LottoLabs]

@napenforcer @albustime 100% I think they naturally complement each other atleast vanilla, you could rice either to be closer to either

[Lotto @LottoLabs]

The new stack is omp w/ 5.5/5.6/fable, hermes w/ local qwen 27b MTP, tailscale

Feels like I’m covered, hermes is more of an assistant, omp is work horse, all off your phone if you need, seamless transfer between setups, one line installs for both

[0xSympathy @SympathyLabs]

@gabu3d_pl @LottoLabs LFG. x8/x8.
@LottoLabs how do you delete runs? agent keeps messing up metadata

[Lotto @LottoLabs]

Let’s goooo

[Campy @c4mpy]

2 weeks later and we're up and running ✅

Dual RTX 5060 Ti 16GBs for < $2k

Thank you @LottoLabs for your help and guidance 🤝

[Lotto @LottoLabs]

@Tech2Wild Pcie4 x8x8 then? I wish there were cheap n links like why aren’t there flexible n links???

[Lotto @LottoLabs]

@Tech2Wild How’s your 3090s configured? Nvlink?

[Lotto @LottoLabs]

@SympathyLabs @gabu3d_pl In your dashboard there’s delete, I’ll double check thought I pushed delete via api too

[Lotto @LottoLabs]

Bros let’s go

TB2.1 style evals will be live soon

Much longer than expected doing e2e testing

A big write up on evals is coming, man is there a lot people don’t see what’s happening behind the scenes

[Lotto @LottoLabs]

@codecovenant I should maybe have it more explicit in the cli to pull that information

[Lotto @LottoLabs]

1800 user accounts on https://t.co/2hx2V3euUv btw

[Lotto @LottoLabs]

@codecovenant There’s already fields for both loaded context and kv cache prefill amount when the benchmarks taken

[Lotto @LottoLabs]

Thank god for competition and open source

[Claude @claudeai]

We're extending Claude Fable 5 access on all paid plans, as well as keeping Claude Code’s weekly rate limits 50% higher, through July 19.

[Lotto @LottoLabs]

This has been my limited experience so far

[Roy @usr_bin_roygbiv]

I've seen multiple examples from myself and others i've spoken with of 5.6 doing things which previously would have taken entire companies or departments of people years of work just a year or two ago in one shot.

People have no idea how large the skill and knowledge gap now is.

[Gabu @gabu3d_pl]

Keep on scrolling, just 2x3090 running NVFP4 35b@ 200t/s ;)))
Is it fast @LottoLabs ?

[Kai Kuspa @kaikuspa]

Oh, speculative decoding is WAY faster 😂 @LottoLabs

[Lotto @LottoLabs]

Word on the street is new qwen open source models are coming out but they’re skipping 3.7

So it’ll be 3.8 or 4 later this year

Sweet sweet 27b

[Lotto @LottoLabs]

Real cool initiative from some reputable names here.

They also run benchmarks on open source abligated models to see if guardrails are effecting leanings

Very cool lm judge system too check it out

[Neutrality Project @neutralityorg]

We're launching The Neutrality Project 🚀, an independent, open-source research collective making AI influence measurable.

AI models answer billions of questions daily. They all come from somewhere: they validate certain options, treat particular middles as reasonable, and leave things out. That influence has been invisible until now.

Founded by @CardilloSamuel, @kaiostephens and @DJLougen, with the help of early contributor @andrewgzavalag1

[Lotto @LottoLabs]

@keennay At a certain point it’s like any other tool, sticking with what you know works faster and easier than learning something new 😭