Nirant

trying my best to escape the permanent underclass with friends

[Nirant @NirantK]

@bijald cc @noheeriye

[Nirant @NirantK]

Someone just spent $450M to get a promotion and create near zero 3 Year terminal value

[Zach Moskow @zachmoskow]

CRAZIEST AI story of the week.

Thomson Reuters just launched an LLM called Thomson.

It starts with Qwen3.5-397B, then is continually trained on 175 years of proprietary legal, tax, accounting, and news data.

They spent $40M on it.

Two years ago, Thomson Reuters bought a pre-revenue legal-AI startup called Safe Sign.

That team built Thomson.

On Thomson Reuters’ published benchmarks, they say it is comparable to Claude Opus 4.8 and ahead of GPT-5.5, Claude Sonnet 5, and Gemini 3.1 Pro overall.

They have used less than 10% of their data collection so far.

We’re going to see this a lot more often.

[Nirant @NirantK]

@Ary_Mulchandani @usefaff faff solves approximately for a bad VA
instinct is zero-setup hermes agent

the difference between a car and pick up truck,
they move different things

both solve for adjacent use cases

[Nirant @NirantK]

poke/instinct will go the clubhouse way for very similar reasons or breakout into spotify space

🇮🇳 investors should speak to @usefaff and what they've learnt

[Nirant @NirantK]

@sidhusmart @usefaff faff is slightly different and more India-friendly pricing at around 1500 INR/mo

I tried it twice but the experience was a mixed bag,
it's rapidly improving though and the unit is "tasks", which is interesting

[Nirant @NirantK]

@kunksed My DJ is called Claude

[Nirant @NirantK]

@mohakmangal @pravah_ai Great idea and looks like a solid data input too.
And @saranormous strikes again!

The RMSE plots are fab, I was going to make them myself 😂

[Nirant @NirantK]

@Ary_Mulchandani welcome to the "dark" side!

happy to recco pour over beans if you'd like

[Nirant @NirantK]

@tech_optimist you're right. I'll be more careful of when to be consise and when to be verbose going forward. I should have been more careful in my past responses to you.

[Nirant @NirantK]

@bijald @Sanyam0605 paryushan meal plans? there's always a tarla dalal answer https://t.co/EgCGVEFCqf

[Nirant @NirantK]

@Sanyam0605 @bijald do what works for you!

[Nirant @NirantK]

Yes, though as I get older I'd have preferred to just do 1, even at the cost of everything else.

Like why would you not work 18 hour days and take on debt to make your business work, makes no sense?

On happiness too, why would you not take a job where you can go be with your parents/wife/kids 6-8 waking hours?

Just picking 1 would make life a lot simpler and mind clearer and the heart lighter

[Nirant @NirantK]

Seasons.

[Ankit @ankitkr0]

it's absurd how intentional you have to be if you want to be happy, healthy and rich all at once.

[Nirant @NirantK]

reposting this because I'd accidentally left the friend's phone number in

but heyy don't mind deleting the previous one!

went viral on WA!

[Nirant @NirantK]

@nishanth_who RAM prices

[Nirant @NirantK]

@bijald This is/was not the case till pre-LLMs. That has hurt the competition and generosity alike.

[Nirant @NirantK]

@ashwingupta some work around pptx and pdf, specifics not general enough to be widely useful

[Nirant @NirantK]

Legit alpha, backed by real $ and time spent for SaaS companies selling agents:

Deepseek Harness can work with any LLM and is absolutely god-tier at cost-perf iterations with repetitive, iterative workloads

Staff 1-2 engineer+PM teams and you'll have an entirely market segment open up

[Nirant @NirantK]

@PradyuPrasad both?