push past limits | real niu for @Gradient_HQ | part time troll | full time janitor | professional edging specialist
edge to edge
Joined December 2024
- Tweets7.3K
- Following669
- Followers2.3K
- Likes18.3K
your provider is slow on serving GPT-6.1 Sol? Try it on @commonstack_ai
OpenAI's best cost efficient model for heavy duty agentic tasks, commonstack offers it at stable production even under changing conditions
build here: commonstack.ai/model-library…
Elon just like me fr
Pretty much.
As of the past few weeks, debt markets have almost entirely shut off access to money for any company that doesn’t have substantial revenue independent of AI data centers.
SpaceX closed $40B of debt for GPUs, but we’re investment grade and have Starlink & launch revenue.
That means the 200+ neoclouds are going to see guppy-level death and consolidation. Some will make it through, but not many.
agents running open model inferences keeping the GPUs brring
Zooming out, GPUs are still renting ahead of a standard 6-year depreciation schedule
Rental price vs. schedule:
- H200: +143%
- B200: +111%
- RTX PRO 6000: +73%
- H100: +54%
- A100: +9%
Most of the gap opened up after April, when token demand took off
Charts of the Week: a16z.news/p/charts-of-the-we…
AI is the only industry right now where you can exist for a year and raise a $1 billion Series A at $6 billion valuation
Clankers accelerating, Jensen and Lisa has some new customers $NVDA $AMD
I think this is probably the greatest advertisement for founders to come to America.
America is the only place where you can lose $80.7 billion and only make $16.3 billion in the same period and have venture capitalists lining up to get inside you
SCOOP: Anthropic's latest draft S-1 documents show financials results for the first half of the year, gigantic spending plans for the future, and a laundry list of risk factors.
- Anthropic recorded a net loss of $80.7 billion in the first half of the year, tied to accounting adjustments related to previous financings
- Anthropic is set to spend >$600 billion on compute in the coming years, inclusive of cloud agreements and chip leases.
More in @WSJ:
one of my personal favorite breakdowns on zero2repo is how the same model performs on different harnesses with the same tasks.
with gpt 6 astra being the same model on codex and opencode it resolved 3 less tasks between the harnesses:
- codex: 15/17 solved
vs
- opencode: 12/17 solved
how can we resolve small bumps that compound as developments continue? these are research efforts that is continuing as agents handle more more workloads.
@Gradient_HQ has more on the way;
- zero2repo: zero2repo.ai/
- github: github.com/Gradient-Data/Zer…
SWE-bench Verified has become saturated. Seven models scored 95–97% (Vals AI, Sep 2026). But how well can coding agents build complete software projects from scratch?
Zero2Repo is a Gradient Data research initiative building hard, verifiable tasks and training data for whole-project development.
Agents start with a full spec, an interface contract and an empty workspace. They build the entire repo without reference code, then submit it for hidden deterministic testing. Every test must pass for a task to count. No LLM judge is used.
The first release includes 17 tasks derived from real open-source projects across Python, TypeScript, Go, C, C++ and Zig.
Across 18 model–harness configurations, scores range from 1/17 to 15/17. The top 10 span 47 percentage points, compared with 5 points on SWE-bench Verified.
No configuration has completed all 17 tasks. This matters for self-iteration: small unresolved issues can persist and compound as agents continue developing a project. Reliable self-evolution remains an open research challenge.
We’re expanding the task collection and looking for contributors to suggest seed repositories, develop tasks and review requirements and hidden tests.
More cases are coming.
👉 zero2repo.ai
👉 github.com/Gradient-Data/Zer…
rw ./ retweeted
SWE-bench Verified has become saturated. Seven models scored 95–97% (Vals AI, Sep 2026). But how well can coding agents build complete software projects from scratch?
Zero2Repo is a Gradient Data research initiative building hard, verifiable tasks and training data for whole-project development.
Agents start with a full spec, an interface contract and an empty workspace. They build the entire repo without reference code, then submit it for hidden deterministic testing. Every test must pass for a task to count. No LLM judge is used.
The first release includes 17 tasks derived from real open-source projects across Python, TypeScript, Go, C, C++ and Zig.
Across 18 model–harness configurations, scores range from 1/17 to 15/17. The top 10 span 47 percentage points, compared with 5 points on SWE-bench Verified.
No configuration has completed all 17 tasks. This matters for self-iteration: small unresolved issues can persist and compound as agents continue developing a project. Reliable self-evolution remains an open research challenge.
We’re expanding the task collection and looking for contributors to suggest seed repositories, develop tasks and review requirements and hidden tests.
More cases are coming.
👉 zero2repo.ai
👉 github.com/Gradient-Data/Zer…
inference and agents under open models go brrrrrrr
🚀 DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.
🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
The world is our post training lab.
Gradient Data - gradientdata.ai
Zero2Repo - zero2repo.ai
We are excited to showcase Gradient Data!
After building Echo and Logits, @Gradient_HQ expanding into verifiable environments, real-world trajectories, and deterministic evaluation systems for more capable and reliable agents.
Zero2Repo is the latest example:
Coding agents must turn a product spec + interface contract into a complete working repository from an empty workspace.
Scoring is binary via hidden deterministic tests, no LLM judges. Current top results reach 15/17 across 17 tasks in multiple languages.
Over the past year, we’ve worked on Echo and Logits to make AI post-training more efficient and accessible.
Gradient Data now extends that work with verifiable environments, real-world trajectories and evaluation systems for more capable and reliable agents.
gradientdata.ai
Build with dozens of other frontier models here: commonstack.ai/model-library
Claude Sonnet 5.5 is live now on @commonstack_ai
Anthropic's new faster, lower-cost model designed for well-scoped everyday tasks, bug fixes, and polished documents. Generating output 30%+ faster while requiring fewer tokens per task, it brings a sharp eye for design, advanced terminal coding agents, and robust cyber safeguards to your agentic workflows.
Try now: commonstack.ai/model-library…
Gemini 4 Argon is live now on @commonstack_ai
Google's new flagship reasoning model built for complex software engineering, deep mathematical and scientific problem solving, and autonomous multi-step agents. Built with a unified 1M-token context window, and robust safety architecture, it delivers frontier performance across robotics. deep coding, enterprise finance/legal workflows, and cybersecurity defense.
Try now: commonstack.ai/model-library…
Build with dozens of other frontier models here: commonstack.ai/model-library
Gemini 4 Argon is live now on @commonstack_ai
Google's new flagship reasoning model built for complex software engineering, deep mathematical and scientific problem solving, and autonomous multi-step agents. Built with a unified 1M-token context window, and robust safety architecture, it delivers frontier performance across robotics. deep coding, enterprise finance/legal workflows, and cybersecurity defense.
Try now: commonstack.ai/model-library…