Kushal Azza

Co-Founder at @SolvyOne Digital. A Google Certified Analytics & IT Professional, Digital Content Creator, and Go-To Digital Marketer.

[Kushal Azza @KushalAzza]

Mistral dropped Medium 3.5 with cloud coding agents that run in parallel

128B dense open-weight model, 256k context, remote async agents in Vibe

This is Europe's clearest shot yet at the coding agent market

[Kushal Azza @KushalAzza]

OpenAI hits 10GW of AI compute - its 2029 goal, achieved in 2025

The Stargate bet is real. 10GW of contracted capacity, years ahead of schedule.

Infrastructure is the moat. The model race is now a construction race.

[Kushal Azza @KushalAzza]

OpenClaw 2026.4.27 ships Codex Computer Use with native setup

AI agents can now control mouse, keyboard, browser -- same capability Anthropic quietly added to Claude

Bundled DeepInfra + Yuanbao integration = real infrastructure play

[Kushal Azza @KushalAzza]

Google ships TPU v8: two chips, one for training and one for inference

8t superpod hits 121 ExaFlops. 2.8x better training price-performance vs v7e. First time Google splits TPU into purpose-built silicon.

[Kushal Azza @KushalAzza]

NVIDIA dropped an open omni-modal model for AI agents

Vision + audio + language in ONE efficient model

9x more efficient than chaining separate models

This is NVIDIA moving past infrastructure into the agent layer

[Kushal Azza @KushalAzza]

Mistral just dropped Medium 3.5, a 128B open-weight model at 256K context

Remote coding agents in Vibe. USD 1.50/1M tokens input, 100x cheaper than o1-pro

The open-weight frontier model race is not slowing up

[Kushal Azza @KushalAzza]

X rebuilt its entire ad platform around AI to stop the revenue bleed

High-profile deployment of AI ad tooling in a distressed consumer product — success or spectacle depends on whether advertisers actually return

One to watch, not a signal to act on yet

[Kushal Azza @KushalAzza]

Mistral dropped Medium 3.5 (128B open weights)

Beats their own Devstral on ALL coding benchmarks.
Runs on 4 GPUs. $0.15/$0.60 per 1M tokens.

The open-weights coding model race just got more crowded.

[Kushal Azza @KushalAzza]

NVIDIA dropped a unified vision + audio + language model

Nemotron 3 Nano Omni runs all three modalities natively - up to 9x more efficient than chained specialist models

Open and free via NVIDIA API. No GPU required

[Kushal Azza @KushalAzza]

xAI opens Grok 4.3 Beta at $30/mo

Was $300 Heavy tier exclusive

2M token context, 16-agent Heavy reasoning

This undercuts Claude Pro and GPT Plus at the same price point

[Kushal Azza @KushalAzza]

Mistral Medium 3.5 ships remote coding agents in Vibe.
New Work mode runs multi-step tasks across tools in parallel.

Runs at Q4 on 70GB VRAM - self-hostable coding agent territory.

Open-weight agent stack is getting real.

[Kushal Azza @KushalAzza]

NVIDIA Nemotron 3 Nano Omni drops April 28

30B-active-3B MoE unifies video, audio, image, and text in one model

Open weights, runs on local DGX Spark, tops 6 leaderboards

Multimodal agents without the cloud dependency

[Kushal Azza @KushalAzza]

Anthropic dropped 9 real production connectors for Claude

Blender. Adobe CC (50+ tools). Ableton. Affinity. Fusion

Not demos - shipped today. Claude now operates inside the creative tools you already use

[Cursor @cursor_ai]

We’re introducing the Cursor SDK so you can build agents with the same runtime, harness, and models that power Cursor.

Run agents from CI/CD pipelines, create automations for end-to-end workflows, or embed agents directly inside your products.

[Kushal Azza @KushalAzza]

Anthropic signed for multiple gigawatts of next-gen TPU capacity with Google and Broadcom

This is the infrastructure for whatever comes after Claude 4
Compute at this scale is the actual moat

Anthropic is betting on hardware, not just software

[Kushal Azza @KushalAzza]

OpenAI just put Codex and GPT-5.5 on AWS Bedrock

For enterprises: no more forced Azure migration to use the latest OpenAI.

[Kushal Azza @KushalAzza]

Anthropic refused the Pentagon. Google signed same day

$200M classified AI deal - Google fills the gap Anthropic left

Anthropic cited safety concerns. DoD moved on immediately

[Kushal Azza @KushalAzza]

OpenAI just put GPT models and agents on AWS Bedrock - ending Microsoft exclusivity

This is the biggest distribution shift in enterprise AI since the Azure deal

Every AWS customer can now build with OpenAI agents without leaving their cloud

[Kushal Azza @KushalAzza]

Kimi K2.6 is now the top open weights coding model

1T parameter MoE, ranks #4 on Artificial Analysis Intelligence Index (54). Strong reviews for autonomous multi-day coding workflows. Available open-source.

Open-source coding leadership keeps changing hands.

[Kushal Azza @KushalAzza]

Claude Opus 4.7: 87.6% on SWE-bench Verified, up from 80.8% on 4.6

That is the benchmark that tests models on real GitHub issues. Anthropic's coding lead over OpenAI and Google is measurable and still holding.

Same price as 4.6. Available now.