Shibi Soundar

[Shyam Siddarth @ShyamSiddarth]

Chennai has 7000 bus stops, 3800 buses.

6 member operation team and 20 day timeline, everything was digitized.

Trained atleast 4000 conductors to go digital.

Its been a year since the launch of Chennai One,

- 1.5 Lac tickets are purchased on a daily basis.
- Monthly, 68k out of 90k bus passes are digitally renewed.

Beautifully designed & executed by a 30 member team (in pic) from @nammayatri @timelessco @Chennai_One
@CumtaOfficial
_
@TamilNaduInfra @UpdatesChennai

[Shibi Soundar @ShibiSoundar]

@DeepakVisva Hope cbe unlinke cities also considers livability and urban design atleast inside private properties and design and develop. Not just growth and economic development but also sustainable and livability in a nice way.

[Shibi Soundar @ShibiSoundar]

@Sanjay_Kasturi @DeepakVisva Had same question

[Shibi Soundar @ShibiSoundar]

@DamonZumbroegel @svembu trying some traditional and cultural aesthetics in their buildings and office designs in @zoho

@DeepakVisva associate with vernacular architecture for his #Vaanalaya project.

[Shibi Soundar @ShibiSoundar]

@svembu Oracle reports 22% rev growth and 121% cloud infra revenue growth in the latest quarter — described as "exceptional" results. spent $1.84 billion on severance in fiscal26. Negative cash flow of $5.4 billion last quarter. $92 billion in AI capex. $2.8 billion restructuring budget.

[Shibi Soundar @ShibiSoundar]

'Today is your last working day': Oracle begins new round of layoffs; here's what the shocking 6 am email says - BusinessToday https://t.co/sveAjGdBur

@svembu

As you pointed recently, the numbers and the narrative don't make sense. It's not poor performance but the AI capex

[Chandru Rajendran @chandruorg]

I received an invite from the Ministry of Finance, and I thought it was spam.

I looked into the email after 15 mins

It said, "Chief Economic Advisor (CEA) of India wants to talk with you"

I still couldn't get it

There was a landline number mentioned in the email. I called up and asked if this was a conference or something I was being invited to.

They said the CEA wants to talk personally and hear the story. And could I come to Delhi.

I said yes. He also wanted all the economists in his office to be part of it.

This is the team that puts together the Economic Survey and the country's economic policies every year.

I flew to Delhi and met the CEA and his team.

I shared my journey into deep-tech, drone policies and Indian manufacturing, in the context of my previous drone startup Airdonex and my current startup @ProcurioIndia

They listened carefully to every challenge we have run into, and were genuinely interested in what founders need from policy.

We concluded with explaining what we are building at PROCURIO, that we are on a mission to set a new benchmark for how anyone can source and manufacture from India.

[Shibi Soundar @ShibiSoundar]

@Tijori1 @svembu your key metric as suggestion to be invested in R&D.

[Shibi Soundar @ShibiSoundar]

@sidddh0 @svembu Hi Siddhanth,
You learn by watching them, being with them, observing what and how they do and also interacting with them. A mentor doesn't teach. One cannot teach all they learnt simply. It's an experience. That's mentorship.

I hope you get a chance to be in Zoho and meet him.

[Sawyer Merritt @SawyerMerritt]

New video of Tesla's revolutionary unboxed manufacturing process for the Cybercab from the launch event.

The production line is designed so that one Cybercab can roll off the line in under 10 seconds, with a long term cycle time goal of 5 seconds, compared to ~34 seconds for a Model Y.

[Shibi Soundar @ShibiSoundar]

@Normal_2610 @svembu owned by Zeiss foundation

[Shibi Soundar @ShibiSoundar]

@zilvestro @AleksDoesCode I can highly advise https://t.co/pvVBciJuJl , it's 34D licensed broker. They specialise for indian expats. Their health insurance expert #Nina is one of the best.

I would say go through a broker than going direct or through an agent. There is a difference on who is on ur side.

[Shibi Soundar @ShibiSoundar]

https://t.co/XcckgCHspE

@svembu @PeterSchiff @KobeissiLetter @Schuldensuehner @Normal_2610

[Shibi Soundar @ShibiSoundar]

@tarunsmehta Th se are really interesting details

[Shibi Soundar @ShibiSoundar]

@mayukh_panja Hi Mayukhy also in Germany. Came to study.

Would like to connect as I see you post more relevant to my thoughts and discussion and question with this living in 2 worlds.

Letme know. Thanks

[Shibi Soundar @ShibiSoundar]

@Normal_2610 @svembu @ZohoSchools @prakdadlani

[Shibi Soundar @ShibiSoundar]

@Normal_2610 @svembu @ZohoSchools fyi

[Shibi Soundar @ShibiSoundar]

@dineshpaii @Rainmatterin #bannariammaninstitute

[Shibi Soundar @ShibiSoundar]

"What in 2023 required paying OpenAI for every token now runs for free, locally, and with total privacy, on Apple's cheapest desktop computer."
@svembu as you said - apple didn't invest or follow the bandwagon. smart money went to their 2 mm M6 chip

[Martin Varsavsky @martinvars]

Apple acaba de presentar el M6, su primer chip de 2 nanómetros. Trae una CPU de 12 núcleos, GPU de 12 núcleos con Neural Accelerators y un Dual Neural Engine de 16 núcleos que duplica la potencia de cálculo de IA de la generación anterior. La GPU rinde casi un 30 por ciento más en cargas de IA que el M5 y hasta 4 veces más que el M4 al que sustituye.

Lo interesante es lo que ya cabe en sus 32 GB de memoria unificada. Con MLX o llama.cpp, un Mac mini de 899 dólares ejecuta gpt-oss-20b, el modelo abierto de OpenAI que rinde al nivel de o3-mini, el mismo que hace poco solo existía en la nube de pago. Ejecuta Qwen3.6 27B, que resuelve el 77 por ciento de SWE-bench Verified, un resultado de programación que supera con claridad lo que ofrecía GPT-4 cuando era el modelo más avanzado del mundo y costaba una fortuna por API. Y ejecuta DeepSeek R1 32B para razonamiento o Gemma 27B con visión y 128K de contexto, todo cuantizado en 4 bits y todo sin enviar un solo dato fuera de tu escritorio.

Conviene saber una cosa: 32 GB es el techo del M6. Quien quiera más memoria en el Mac mini tiene que bajar a la generación anterior con el M5 Pro, así que el chip más nuevo de Apple es también el que menos RAM admite. Eso deja fuera la clase 70B y marca con claridad para quién es esta máquina: modelos medianos cuantizados, agentes privados y fine-tuning ligero, no IA pesada.

La conclusión sigue siendo muy especial. Lo que en 2023 exigía pagar a OpenAI por cada token hoy corre gratis, en local y con total privacidad, en el ordenador de sobremesa más barato de Apple.

Las reservas se abren hoy y las máquinas llegan el 22 de septiembre. La frontera de la IA de hace dos años ahora cuesta $899 dólares y cabe en la palma de la mano.

[Shibi Soundar @ShibiSoundar]

@VishakhRanotra Ideas come from everwhere and spark anytime. It's how we capture it and execute it. This is what is entrepreneur spark.

Great example ji.