Tech & AI leader, Board Director, Investor, Art & Design lover, Urban Farmer with an alter ego @CrazyPlantLadee https://xitter.cf/t.co/kmW6da0R03
Joined May 2009
- Tweets5K
- Following2.2K
- Followers24.4K
- Likes8.9K
Did you know?
What you hear can shape your metabolic health.
Environmental noise triggers stress, increasing cortisol levels that can disrupt insulin response.#geniehealth
Maya Hari retweeted
BREAKING: Anthropic reveals Claude Mythos discovered post-quantum cryptographic weaknesses that had evaded years of human expert review.
The joy of a lazy Sunday and losing yourself in the pages of a delightful book @ChimamandaReal
Lost productivity to extreme weather will increasingly be a drag on country's GDPs. The importance of energy grid is certainly key. Equally, countries need to tackle the issue upstream-reducing emissions & decarbonization to prevent further climate worsening. #bendthecurve
This quoted post is unavailable.
Maya Hari retweeted
Eli Lilly to sign $2bn deal for AI drug development with Hong Kong biotech ft.trib.al/lRujYfm
Maya Hari retweeted
While social media is polarising, evidence suggests AI may nudge people towards the centre.
This holds true of all studied models. Grok is more right-leaning than other models, but also has depolarising effects.
By @jburnmurdoch.
Maya Hari retweeted
If you're an AI startup in India, renting processing power from the government to train your model costs about $0.7 per hour. The same hardware on Amazon Web Services costs $3.7. On Microsoft Azure, $6.6. The Indian government is subsidizing AI infrastructure at rates that would make most Western startups do a double-take.
I read all 26 pages of the white paper this tweet links to. The numbers inside are wild.
The IndiaAI Mission has a budget of about $1.2 billion over five years, approved in March 2024. Almost half of that, roughly $500 million, goes straight to building the processing power AI companies need to train their models. The original plan was to deploy 10,000 processors. By December 2025, they had 38,000 running. 3.8x what they promised.
A government open call in January 2025 pulled 506 proposals. The four startups picked first were Sarvam AI, Soket AI, Gnani AI, and Gan AI. Eight more were added by September. India now has 12 separate teams building AI models, ranging from tiny ones for basic chatbots to massive ones rivaling those from the US and China. They cover language, voice, vision, medical diagnosis, material science, and even brain-computer interfaces.
The one I keep coming back to is Sarvam AI. They raised $41 million from Lightspeed, Peak XV, and Khosla Ventures. In May 2025, they released a model built on top of a French AI system (Mistral Small) and customized for Indian languages. It got roasted online. Critics said it was a foreign model in Indian clothing. So they went back and built Sarvam-105B completely from scratch, using Indian hardware under the government mission. It outperformed China's DeepSeek-R1 on certain tests, even though it was a model six times larger. Both were released for anyone to download and use in March 2026.
There's something else buried in the paper I haven't seen another country try at this scale. India is building a copyright system specifically for AI training data. Under a December 2025 government proposal, AI companies can train their models on any copyrighted content they can legally access, books, articles, music, anything. Creators cannot say no. But the moment an AI product makes money, royalties are collected by a centralized government body and distributed back to creators. Singapore allows AI companies to use content without payment. China requires strict consent before training. India is trying a middle path, and publishers are already calling it forced participation.
Stanford's AI Vibrancy Index, which measures a country's overall AI strength across research, talent, infrastructure, and investment, ranked India third globally in 2025. Up from seventh in 2023. But the actual scores tell you how far the gap still is: US at 79, China at 37, India at 22. And India's $1.2 billion budget sits next to China's $47.5 billion semiconductor fund and Saudi Arabia's $100 billion Project Transcendence.
India is currently spending 40x less than the frontrunners. This white paper is the most detailed public bet yet that smart infrastructure design can close that gap.
𝐀𝐬 𝐩𝐚𝐫𝐭 𝐨𝐟 𝐭𝐡𝐞 𝐨𝐧-𝐠𝐨𝐢𝐧𝐠 𝐀𝐈 𝐏𝐨𝐥𝐢𝐜𝐲 𝐖𝐡𝐢𝐭𝐞 𝐏𝐚𝐩𝐞𝐫 𝐒𝐞𝐫𝐢𝐞𝐬, 𝐭𝐡𝐞 𝐎𝐟𝐟𝐢𝐜𝐞 𝐨𝐟 𝐭𝐡𝐞 𝐏𝐫𝐢𝐧𝐜𝐢𝐩𝐚𝐥 𝐒𝐜𝐢𝐞𝐧𝐭𝐢𝐟𝐢𝐜 𝐀𝐝𝐯𝐢𝐬𝐞𝐫 𝐭𝐨 𝐭𝐡𝐞 𝐆𝐨𝐯𝐞𝐫𝐧𝐦𝐞𝐧𝐭 𝐨𝐟 𝐈𝐧𝐝𝐢𝐚 𝐫𝐞𝐥𝐞𝐚𝐬𝐞𝐬 𝐚 𝐰𝐡𝐢𝐭𝐞 𝐩𝐚𝐩𝐞𝐫 𝐨𝐧 “𝐀𝐝𝐯𝐚𝐧𝐜𝐢𝐧𝐠 𝐈𝐧𝐝𝐢𝐠𝐞𝐧𝐨𝐮𝐬 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧 𝐌𝐨𝐝𝐞𝐥𝐬.
The versatility of Foundation Models makes them a critical layer of today’s AI ecosystem and a key area for innovation in India. Therefore, developing indigenous foundation models is a strategic priority. India’s objective is to harness foundation models for inclusive growth and public good, while ensuring they are governed in a manner consistent with the country’s values, legal framework, and security interests.
This white paper provides an understanding of India’s approach to advancing indigenous foundation models through public–private collaboration and to governing these systems that support trust, accountability, and responsible adoption.
The White Paper also provides details on India’s approach - which is centred on building indigenous capability across the foundation-model stack. Rather than relying on a single model, India is developing an ecosystem that combines
(i) shared compute access,
(ii) India-centric data and model repositories, and
(iii) multiple model-building efforts across text, speech, multimodal, and sectoral systems.
Read the White Paper here: psa.gov.in/CMS/web/sites/def…
When the most unusual plants flower, its beathtaking #motherinlawstongue #snakeplant
Maya Hari retweeted
When you see how people use Claude Code/Codex/etc it becomes clear that managing agents is really a management problem
Can you specify goals? Can you provide context? Can you divide up tasks? Can you give feedback?
These are teachable skills. Also UIs need to support management
Maya Hari retweeted
New Stanford–Harvard study: widely used AI models can cause clinical harm in up to ~22% of cases.
Top models still produce ~12–15 severe harms per 100 cases; worst models exceed 40.
Most harm comes from omissions, not bad advice (≈77% of errors; eg failing to order a critical test).
Some key takeaways:
• Best models outperform generalist physicians using conventional resources
• Clinical safety doesn’t correlate with model size, recency, “reasoning-mode,” or popular benchmarks (eg MedQA; only moderate correlation r ≈ 0.61–0.64)
• Multi-agent + RAG meaningfully reduce harm: heterogeneous 3-model ensembles show ~6× higher odds of top-quartile safety vs solo models
Maya Hari retweeted
My top takeaways from @jasonlk:
1. AI is enabling smaller, leaner teams to punch above their weight. Companies that used to need 10 salespeople to hit a revenue target are now achieving the same results with five or six, thanks to AI-powered automations and ever-smarter agents.
2. AI agents can match human salespeople’s performance when properly trained. SaaStr replaced their entire 10-person go-to-market team with 20 AI agents managed by 1.2 humans, while maintaining the same revenue levels. “It’s not better; it’s not worse. But it’s so much more efficient, and it scales like software scales.”
3. GTM efficiency is now a competitive moat, and AI is the primary lever for achieving it. In a market where growth at all costs is no longer rewarded, Jason sees AI as the way for B2B companies to maintain or accelerate revenue growth while keeping sales and marketing costs flat. Companies that master this balance will be able to outlast and outcompete.
4. The mediocre middle is most at risk. “AI is replacing the jobs people don’t want to do today. And it is displacing the mid-pack and the mediocre—their jobs are very much at risk.” The best salespeople will build their AI superpowers, but those who just show up and don’t truly understand their job are going to be quickly replaced by AI.
5. The traditional advice to hire your first sales rep at $1M ARR is becoming obsolete in an AI-first world. Jason is observing portfolio companies now reaching $5M, even $10M, ARR with founder-led sales supplemented by AI tooling, automation, and highly efficient inbound funnels. This doesn’t mean sales teams disappear, but it does mean the timing, structure, and skill set required are fundamentally different than they were 18 months ago.
6. The traditional playbook of hiring large SDR teams is over as AI tools take over prospecting and initial outreach. Jason has built AI-driven outbound engines that can personalize emails, qualify leads, and book meetings at a fraction of the cost of human SDRs. Founders who continue to scale traditional SDR orgs without testing AI alternatives risk overspending on a function that’s being rapidly commoditized.
7. The bar for cold outbound has skyrocketed—generic spray-and-pray is now filtered out by AI gatekeepers on the buyer side. The companies still seeing success with outbound are using hyper-personalized, research-driven sequences that feel one-to-one, often powered by AI that scrapes LinkedIn, monitors job changes, and tailors messaging to the individual. If your outbound isn’t informed by real-time signals and contextual relevance, it’s already in the spam folder.
8. The biggest mistake founders are making right now is waiting for AI to stabilize before adopting it—by the time the dust settles, the gap between early adopters and laggards will be insurmountable. Jason is urging his portfolio to experiment aggressively, fail fast, and build institutional knowledge around what works. The companies that figure out AI-powered sales will have a 12- to 18-month head start that compounds into a durable competitive moat. The time to move is now, not when best practices emerge.
.@jasonlk used to have 10 salespeople. Now he has 10 empty desks—each one labeled with an AI agent's name.
His sales team went from 10 humans → 1.2 humans + 20 AI agents. Same revenue.
In his return to the podcast, we discuss:
🔸 Why mid-tier SDRs/BDRs will be "extinct" within a year
🔸 Where to start integrating AI into your sales motion
🔸 The specific AI tools that have worked best (and which flopped)
🔸 His 2026 predictions for GTM and sales
🔸 How to become "hyper-employable" in the age of AI
Listen now 👇
• YouTube: youtu.be/I-R1bc1rlFs
• Spotify: open.spotify.com/episode/5lt…
• Apple: podcasts.apple.com/us/podcas…
Thank you to our wonderful sponsors for supporting the podcast:
🏆 @DeveloperXM — The developer intelligence platform designed by leading researchers: getdx.com/lenny
🏆 @vercel — Your collaborative AI assistant to design, iterate, and scale full-stack applications for the web: vercel.com/lennyspodcast
🏆 @datadoghq — Now home to Eppo, the leading experimentation and feature flagging platform: datadoghq.com/lenny
Maya Hari retweeted
I've never felt this much behind as a programmer. The profession is being dramatically refactored as the bits contributed by the programmer are increasingly sparse and between. I have a sense that I could be 10X more powerful if I just properly string together what has become available over the last ~year and a failure to claim the boost feels decidedly like skill issue. There's a new programmable layer of abstraction to master (in addition to the usual layers below) involving agents, subagents, their prompts, contexts, memory, modes, permissions, tools, plugins, skills, hooks, MCP, LSP, slash commands, workflows, IDE integrations, and a need to build an all-encompassing mental model for strengths and pitfalls of fundamentally stochastic, fallible, unintelligible and changing entities suddenly intermingled with what used to be good old fashioned engineering. Clearly some powerful alien tool was handed around except it comes with no manual and everyone has to figure out how to hold it and operate it, while the resulting magnitude 9 earthquake is rocking the profession. Roll up your sleeves to not fall behind.