Sid (@sidgraph) header
Sid (@sidgraph) profile picture

Energy, Taste, and Intelligence Cofo @Basethesislabs, e/acc @GenesisAILabs

Earth
Joined September 2019
Photo by Sid (@sidgraph)Photo by Sid (@sidgraph)Photo by Sid (@sidgraph)Photo by Sid (@sidgraph)Photo by Sid (@sidgraph)Photo by Sid (@sidgraph)Photo by Sid (@sidgraph)Photo by Sid (@sidgraph)Photo by Sid (@sidgraph)Photo by Sid (@sidgraph)
Have been reading about state of venture capital in india. Some notes: docs.google.com/document/d/1…
1
227
Enterprise AI diffusion in wealth, finance, and venture hasn't hit the P&L. Asset managers: tech spend +8.9% CAGR over 5 yrs, margins down 3–5% Wealth managers: AuM +13%, revenue +7.1% Private credit: AI adoption ~50%, ~60% still monitor portfolios on spreadsheets Spend is going into tools that save time inside existing workflows. It is not going into workflows that raise throughput per expert. Hours saved without more clients served do not show up in margin. What the numbers say across segments: ➡️ The return on an agent scales with the firm’s cost ratio. The thinner the margin, the more each unit of expert capacity is worth. Back-office and CFO services firms have the most to gain. ➡️ In wealth, compensation is the largest cost line. Revenue per RM moves cost-to-income several times more than any ops cut. ➡️ In venture debt, the cost base is already lean. The value is in loss avoidance: catching stress in monthly monitoring while there is still leverage. The metric that matters is net capacity: clients served per expert after review, rework, and vendor support are counted.
We’re building Prime at @Basethesislabs so firms can take on a hell of a lot more clients with the expertise they already have. And finally... finally build and offer the services they’ve never had the time for. @sheikh_javeed12 , @sidgraph , @picardo_ria , @dhimant , @RaveenSastry
2
1
2
268
New article: why headcount no longer sets a professional firm’s delivery capacity. In most firms, delivery is capped by the availability of a few senior experts: partners, RMs, specialists. Their corrections are rarely captured, so the same issues return to them every cycle. Our approach at BaseThesis: • Start with one high-frequency, checkable workflow chosen with the firm • Convert expert corrections into tested rules and reusable workflows • Measure net capacity after review, rework and our own support time teapawt.net/thesisofsarthak/status…
1
29
Biology is hard to generalize. But it is ready to specialize on your data and solve your problem. Your lab already collects plenty of data. What it lacks are the systems that turn it into a learnable signal: state-of-the-art models, environments and evals, built around your problem. LiteFold converts your own lab, into an AI learning environment, so you can recursively self improve for your the target you are working or the molecule you are designing. We call this: Recursive Improvement From Experiment Feedback (RIEF) Try LiteFold today and contact us if you want to integrate LiteFold to your lab.
10
29
4
117
11,680
Replying to @archgen_
@archgen_ is building skyscrapers for semiconductor chips Shrinking transistors is getting harder. Stacking them makes heat harder to manage. We’re building the tools to design chips in three dimensions. We stack transistors vertically, shorten connections, and optimize timing and heat together. Our foundational model for thermal simulation, predicts temperature roughly 5,000× faster than conventional EDA tools. Our placement engine ranked #1 in HRT’s macro-placement challenge We’re working with some of the largest fabs in the world to tape out our first chip in April 2027 archgenlabs.com
99
98
47
545
102,395
Hill climbing bio 🚀🚀
Introducing the new LiteFold. Now generally available. Hill Climb Biology.
2
2
244
Replying to @AnthropicAI
@AnthropicAI S-1 as an economics document 🧵! shorturl.at/28jnd
1
2
198
This economy is being financed in a loop and is correlated with demand compounding. Google, Amazon, and Microsoft: > ≥$114.5B, ≥$110B and ≥$44.9B of compute. > 48% of revenue flows through their marketplaces. AMD sells it >$20B of chips and buys up to $5B of its stock. Even the $72.4B H1 “loss” on convertibles shows up as a gain on partners’ books. Now add ~$853B of mostly fixed commitments against ~90% incremental margins. Those are railroad and telecom economics: huge fixed cost, near-zero marginal cost. That structure ends in 2-3 players, entry tickets in the hundreds of billions, and the #2 forced to compete on price.
1
28
The core risk is a maturity mismatch between revenue and obligations. Revenue: 80% is usage-based and can fall month to month. Contracted backlog (RPO) is $10.8B, ~75% due within 12 months, against a $69B run-rate. Costs: ~$853B of compute commitments, ~$430B non-cancelable. Broadcom TPU leases run 5 years, Google cloud to 2033, Amazon to 2036, the tail to 2045. Product: >80% of Q3 revenue came from models released in the prior 6 months. The revenue base turns over every ~2 quarters, so R&D at ~60% of revenue works as maintenance spend, not optional growth spend. Equity: the reported $1.5-2T target is 22-29x run-rate for a PBC where founders hold 50.1% of votes and a trust elects directors. Public holders get the economics, not control. If revenue per compute $ keeps rising, fixed costs spread over more revenue and margins expand. If it flattens, $132B of cash plus IPO proceeds covers the gap.
6
1/4 Meet Qwenputer: Qwen3.5-4B with a full RISC-V CPU inside its weights, using the model's own Gated DeltaNet layers as RAM. It runs C and MicroPython in its forward passes, no external runtime. On DeepMind math: base 55–90%, base + Python tool 95–100%, Qwenputer 97.5–100%.
305
755
354
6,120
898,103
Dear Prime Minister Ambani, Please accept my humble apologies for not realizing that you are the real boss of India. Naturally, you would prefer to maintain your monopolistic exploitation of the great people of India, but would you nonetheless consider allowing Starlink to compete? There are many areas of India, as there are in all countries, with no Internet access, which denies children the opportunity for education and for small businesses the opportunity to sell their goods to a global market. Starlink would change their lives for the better. Thank you 🙏
101,615
71,790
28,061
548,582
118,851,927
Eyecandy ♥️
hardware is having its moment. we’re building the robots we grew up watching in movies.
1
272
launching soon… @eyecandyrobots
23
39
8
470
30,271
Amazing read! Every firm will soon produce the same work at the same quality on the same day, signatures dissolving into a fluent interchangeable average. We deploy aligned systems inside great professional firms so unique taste, recipe and judgment scale while secret ingredients stay the firm’s own. In our first fifty days inside a working tax advisory firm, systems independently re-derived 430,766 figures and intercepted roughly one in every 156. Intelligence inside a firm should be aligned to it, provable to it, and preserving of it, amplifying the spike rather than averaging it away. Nothing leaves without the responsible partner’s confirmation; we build workpapers so claims are usable by everyone whose decisions count on proof.
Every services firm is about to produce the same work, at the same quality, on the same day Our AI re-derived 430,766 figures; 1 in 156 was about to ship wrong. This peice is for anyone who became excellent in their own way, and intends to stay that way.
1
1
7
868
I was actually enjoying my Saturday 😭
3
210
Sid retweeted
really well written blog on the emerging shift from optimization-limited to rollout-limited frontier RL. what stood out to me was the observation that as reasoning trajectories become longer and more heterogeneous GPU utilization and learner-generator synchronization increasingly dominate training efficiency. one underexplored question is whether capability jumps themselves can be detected through staleness statistics. if old trajectories suddenly become much less useful for learning that may indicate the model has discovered a qualitatively new reasoning strategy. in that sense replay-buffer value decay could become an observable signal of phase transitions in capability acquisition. if you're into async-RL/large-scale training systems this is a really worthwhile read.
New blog! Is frontier asynchronous RL solved? The blog covers Async RL theory and infrastructure, surveying 8 open-weight frontier labs for the algorithmic techniques and systems fixes to handle train-inference mismatch. Also answered: why do current methods still fail at high policy lag? Which methods scale with horizon and compute?
2
27
297
43,990
The open-source protein ML space just got a massive upgrade. Phenomenal work by @anindyadeeps and @try_litefold on dropping the biggest protein data collection on Hugging Face
We have released the biggest protein data collection on Hugging Face, guys! We have been working on this for more than 3 weeks now, starting from curating the raw data, doing a lot of filtering, splitting the datasets, sharding them, and doing a lot of analysis. Everything is summarized in our recent blog post.
1
5
10
2,097
Check this out 👇
open sourcing Marlin-2B 🐟 a tiny VLM to extract structured information from videos Marlin is finetuned for two questions devs want to ask in their videos: what is happening, and when? Best open model in its weight class, competitive with Gemini-2.5-flash at only 2B params 🧵
6
554
Damn — its soo over for open ai 🫡
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
2
341