Prashant Deva

Geek. I like video games and math. Chief guy @

[Prashant Deva @pdeva]

hot take: without @SBF_FTX leading Anthropic's Series B, the world would be a different place.

No Anthropic means:
- Openai has monopoly on frontier AI
- token prices are 10x higher
- coding AI revolution takes lot longer

For a glimpse of that world, see O1 token pricing. Thats OpenAI operating without competition..

[Prashant Deva @pdeva]

google crawler ran into exception indexing granola website and made that the content description...

[Prashant Deva @pdeva]

@Google @antigravity helps to test your auth before deploying...

[Prashant Deva @pdeva]

Antigravity 2.0 failing auth...

[Prashant Deva @pdeva]

i wish cursor folks would release more benchmark results so better comparisons can be made

[Prashant Deva @pdeva]

so automated code reviews via Codex cloud still use gpt-5.3-codex. not gpt-5.5...

[Prashant Deva @pdeva]

actually lowkey impressed github ui doesnt completely crawl on this pr. was expecting it to completely freeze
https://t.co/Xe7VEEVrGx

[Prashant Deva @pdeva]

Cursor Bugbot about to get expensive..

[Prashant Deva @pdeva]

every monday github is down...

[Prashant Deva @pdeva]

best explanation of the deep technical forces driving the xai anthropic deal

[Jukan @COMPUTEX @jukan05]

Why did xAI hand over a 220,000-GPU cluster to Anthropic?

The technical backdrop to xAI's decision to hand Colossus 1 over to Anthropic in its entirety is more interesting than it appears. xAI deployed more than 220,000 NVIDIA GPUs at its Colossus 1 data center in Memphis. Of these, roughly 150,000 are estimated to be H100s, 50,000 H200s, and 20,000 GB200s. In other words, three different generations of silicon are mixed together inside a single cluster — a "heterogeneous architecture."

For distributed training, however, this configuration is close to a disaster, according to engineers familiar with the setup. In distributed training, 100,000 GPUs must finish a single step simultaneously before the cluster can advance to the next one. Even if the GB200s finish their computation first, the remaining 99,999 chips have to wait for the slower H100s — or for any GPU that has hit a stack-related snag — to catch up. This is known as the straggler effect. The 11% GPU utilization rate (MFU: the share of theoretical FLOPs actually realized) at xAI recently reported by The Information can be read as the numerical fallout of this problem. It stands in stark contrast to the 40%-plus MFU figures achieved by Meta and Google.

The problem runs deeper still. As discussed earlier, NVIDIA's NCCL has traditionally been optimized for a ring topology. It works beautifully at the 1,000–10,000 GPU scale, but once you push into the 100,000-unit range, the latency of data traversing the ring once around becomes punishingly long. GPUs need to churn through computations rapidly to keep MFU high, but while they sit waiting endlessly for data to arrive over the network fabric, more than half of the silicon falls into idle. Google sidestepped this bottleneck with its own custom topology (Google's OCS: Apollo/Palomar), but xAI, by my read, has not yet reached that stage.

Layer Blackwell's (GB200) "power smoothing" issue on top, and the picture comes into focus. According to Zeeshan Patel, formerly in charge of multimodal pre-training at xAI, Blackwell GPUs draw power so aggressively that the chip itself includes a hardware feature for smoothing power delivery. xAI's existing software stack, however, was optimized for Hopper and does not understand the characteristics of the new hardware; when it imposes irregular loads on the chip, the silicon physically destructs — literally melts. That means the modeling stack must be rewritten from scratch, which in turn means scaling is far harder than most of us imagine.

Pulling all of this together points to a single conclusion. xAI judged that training frontier models on Colossus 1 simply was not efficient enough to be worthwhile. It therefore moved its own training workloads wholesale onto Colossus 2, built as a 100% Blackwell homogeneous cluster. Colossus 1, on the other hand — whose mixed architecture is far less crippling for inference, which parallelizes more forgivingly — was leased in its entirety to an Anthropic that desperately needed inference capacity.

Many observers point to what looks like a contradiction: Elon Musk poured enormous capital into building Colossus, only to hand the core asset over to a direct competitor in Anthropic. Others read it as xAI capitulating because it is a "middling frontier lab." But these are surface-level reads.

Look at the numbers and a different picture emerges. xAI today holds roughly 550,000+ GPUs in total (on an H100-equivalent performance basis), and Colossus 1 (220,000 units) accounts for only about 40% of the total available capacity. Colossus 2 — built entirely on Blackwell — is already operational and continuing to expand. Elon kept the all-Blackwell homogeneous cluster (Colossus 2) for himself and leased out the older, mixed-generation Colossus 1. In other words, he handed the pain of rewriting the stack — the MFU-11% debacle — to Anthropic, while keeping his own focus on training the next generation of models.

The real point, then, is this. Elon's objective appears to be positioning ahead of the SpaceXAI IPO at a $1.75 trillion valuation, currently floated for as early as June. The narrative SpaceXAI now needs is that xAI — long the "sore finger" — is not merely a research lab burning cash, but a business with a "neo-cloud" model in the mold of AWS, capable of leasing surplus assets at high yields.

From a cost-of-capital perspective, an "AGI cash incinerator" is far less attractive to investors than a "data-center landlord generating cash."

As noted above, the most important detail of the Colossus 1 lease is that it is for inference, not training. Unlike training, inference requires far less tightly synchronized inter-GPU communication. Even when the chips are heterogeneous, the workload parcels out cleanly across them in parallel. The straggler effect — the chief weakness of a mixed cluster — is essentially neutralized for inference workloads.

Furthermore, with Anthropic occupying all 220,000 GPUs as a single tenant, the network-switch jitter (unanticipated latency) that arises under multi-tenancy disappears. The two sides' technical weaknesses end up complementing each other almost exactly.

One insight follows. As a training cluster mixing H100/H200/GB200, Colossus 1 was an asset that could only deliver an MFU of 11%. The moment it was handed over to a single inference customer, however, that asset transformed into a cash-flow asset rented out at roughly $2.60 per GPU-hour (a weighted average of the lease rates across GPU types). For xAI, what was a "cluster from hell" for training has become a "golden goose" minting $5–6 billion in annual revenue when redeployed for inference. Elon's genius, I would argue, lies not in the model but in this asset-rotation structure.

The weight of that $6 billion becomes clearer when set against xAI's income statement. Annualizing xAI's 1Q26 net loss yields roughly $6 billion in losses per year. The $5–6 billion in annual revenue generated by leasing Colossus 1 to Anthropic, in other words, almost perfectly hedges xAI's loss figure. This single deal effectively pulls xAI to break-even.

Heading into the SpaceXAI IPO, this functions as a core line of financial defense. From a cost-of-capital standpoint, if the image shifts from "research lab burning cash" to "infrastructure tollgate stably printing $6 billion a year," the entire tone of the offering can change.

(May 8, 2026, Mirae Asset Securities)

[Prashant Deva @pdeva]

Claude Code:
1. Write in NodeJs to 'move fast'.
2. Needs to run on Bun to get any perf.
3. Acquire Bun cause its too critical now.
4. Rewrite Bun from Zig to Rust to get memory safety.

Codex:
- Start writing in Nodejs.
- Realize mistake.
- Rewrite Codex from scratch in Rust.

[Prashant Deva @pdeva]

its absolutely ridiculous that Opus Fast Mode support 4.6 but not 4.7...

[Prashant Deva @pdeva]

seriously thinking of moving to gitlab at this point #githubOutage

[Prashant Deva @pdeva]

dear @GeminiApp just because input the prompt using voice doesn’t mean I also want the output to be read aloud to me. at least give me a setting to toggle this off. @OfficialLoganK

[Prashant Deva @pdeva]

a Michael Jackson documentary without any Michael Jackson songs... there is just Bad at the end and a minute each of Bille Jean and Thriller.
The rest is just the most boring ass student film of all time. The Michael Jackson songs you want to hear are simply not in this movie.

[Prashant Deva @pdeva]

GPT is now more expensive than Opus..

[Max Weinbach @mweinbach]

GPT 5.5 is 2x the price of GPT 5.4 at $5 per 1M tokens in and $30 per 1M tokens out

[Prashant Deva @pdeva]

@BenjaminDEKR the acquisition is contingent on them building a frontier level model that can compete head to head with Opus. Given that, the price is a steal. Who wouldn’t acquire the next Anthropic for $60B

[Prashant Deva @pdeva]

Seems a lot of people are chalking this to desperation of XAI to catch up. I think it's the opposite.

Cursor was always destined to fold into one of the larger AI labs. They simply don't have the capital needed to train models large enough to compete at the frontier. VC funding can only go so far.

The large labs have had to rely on funding from major corporations to fund their training runs. And those corporations have already made their bets on other leading AI labs. This leaves Cursor in a bit of a no-man's land as other labs will definitely catch up to their product by end of this year.

This deal gives them $10billion of compute to catch up to the frontier.
If they do, the reward is to get acquired for $60B.
If they don't, they are left to a flat-line valuation.

In an alternate reality where they did have enough funding, there would be no reason to train a frontier model only to get acquired at 20% premium from the recent $50B valuation.

Its a win-win for SpaceX, since it has unused compute right now. If Cursor does crack the frontier, then this would be an incredible investment. If it doesnt, no real loss.

[SpaceX @SpaceX]

SpaceXAI and @cursor_ai are now working closely together to create the world’s best coding and knowledge work AI.

The combination of Cursor’s leading product and distribution to expert software engineers with SpaceX’s million H100 equivalent Colossus training supercomputer will allow us to build the world’s most useful models.

Cursor has also given SpaceX the right to acquire Cursor later this year for $60 billion or pay $10 billion for our work together.

[Prashant Deva @pdeva]

kimi 2.6 vs composer 2 (which is based on kimi 2.5) is the comparison i wish someone would do

[Kimi.ai @Kimi_Moonshot]

Meet Kimi K2.6: Advancing Open-Source Coding

🔹Open-source SOTA on HLE w/ tools (54.0), SWE-Bench Pro (58.6), SWE-bench Multilingual (76.7), BrowseComp (83.2), Toolathlon (50.0), Charxiv w/ python(86.7), Math Vision w/ python (93.2)

What's new:
🔹Long-horizon coding - 4,000+ tool calls, over 12 hours of continuous execution, with generalization across languages (Rust, Go, Python) and tasks (frontend, devops, perf optimization).
🔹Motion-rich frontend - Videos in hero sections, WebGL shaders, GSAP + Framer Motion, Three.js 3D.
🔹Agent Swarms, elevated - 300 parallel sub-agents × 4,000 steps per run (up from K2.5's 100 / 1,500). One prompt, 100+ files.
🔹Proactive Agents - K2.6 model powers OpenClaw, Hermes Agent, etc for 24/7 autonomous ops.
🔹Claw Groups (research preview) - bring your own agents, command your friends', bots & humans in the loop.
-
K2.6 is now live on https://t.co/YutVbwktG0 in chat mode and agent mode.
For production-grade coding, pair K2.6 with Kimi Code: https://t.co/uvoSJKyGCY
-
🔗 API: https://t.co/EOZkbOwCN4
🔗 Tech blog: https://t.co/9wWvgIQSS3
🔗 Weights & code: https://t.co/Be0hjs2RTP

[Prashant Deva @pdeva]

apparently Fast Mode is not available for Opus 4.7...