Skip to content

The Monday brief: Week of Oct 19–25, 2026

15 stories from 15 companies

Nvidia and TCV back Higgsfield in Series B extension as it passes $1B run rate

Higgsfield announced that TCV and Nvidia invested in an extension of the Series B it raised in August, led by DST Global. It did not disclose the size of the extension or the valuation. The company said it crossed $1 billion in annualized revenue in September, about 18 months after launch. It reported more than 40 million users, with enterprise as its fastest-growing segment. Bloomberg reported this is Nvidia's first investment in Higgsfield, and the announcement coincided with Jensen Huang's first visit to Kazakhstan, where the company has deep roots.

Generative video is reaching enterprise scale, with brands using it to make many ad variants; expect requests for media-generation pipelines.

Baseten launches an Environment Network of vetted partners to build custom RL environments

Baseten introduced the Baseten Environment Network, a shortlist of audited RL environment companies that will build custom post-training environments from each customer's own data. Partners such as Snorkel AI, Abundant, Proximal, Eigent, micro1, Preference Model, Arcanum and Taste Labs work directly with customers on environments, tasks and rubrics, while Baseten keeps the training and production inference. Baseten says it has done this work in-house for customers like Harvey, Lovable and Gamma and needed more expert teams to scale it. It also plans to open-source a set of frontier-quality environments along with its auditing and training methods. Several partners announced their participation on X the same day.

Teams post-training their own models can now buy custom RL environments as a service while keeping ownership of them, rather than building that expertise in-house.

BFL says FLUX 3 Image Klein now ranks #1 among open-weight text-to-image models

Black Forest Labs said FLUX 3 Image Klein ranks first in text-to-image among open-weight models on third-party benchmarks, and that its own testing shows the model is strongest at precise image editing. Posts tracking the Artificial Analysis leaderboard noted the 5.6B-parameter model leads Qwen-Image 2.1 by a single point. Some users welcomed the result and praised its hand rendering, while others were tired of rotating 'number one' claims and asked BFL to release open weights for FLUX 3 Dev editing instead. The model, released October 15, can be fine-tuned with LoRA and runs on consumer GPUs.

A small open-weight image model leading third-party boards makes self-hosted image editing a realistic option over paid APIs.

OpenRouter lists all Poolside Laguna 2.1 endpoints, paid and free, as going away Oct 31

A day after Laguna S 2.1 showed up as a free model in Cline, OpenRouter's model pages now mark Poolside's Laguna S 2.1 and Laguna XS 2.1 as going away on October 31, 2026. This applies to both the free and paid variants, not just the promotional free tier. Posts on X tracking free model quotas flagged the end date, while others kept sharing Cline's free model list that includes Laguna S 2.1, a 118B-parameter coding model with 8B active. Poolside has not publicly explained the change, so it is unclear whether the models are moving elsewhere or being retired.

Teams trialing Laguna via OpenRouter should plan for the Oct 31 cutoff and have a fallback coding model ready.

Robinhood Ventures discloses ~$10M stake in Cognition as Devin passes $1B run rate

Robinhood Ventures Fund I, a publicly traded closed-end fund, said it closed about $10 million of investments in Cognition. According to Fortune, the stake was part of Cognition's September Series E, which raised more than $2 billion at a $48 billion valuation; Robinhood is only now disclosing its participation. Robinhood's release says Cognition recently crossed $1 billion in annualized revenue run rate and lists Goldman Sachs, Citi and Nubank as Devin customers. Robinhood had already invested in OpenAI, and its fund head said AI coding is not a winner-take-all market.

A $1B run rate and large-bank deployments show autonomous coding agents are now mainstream enterprise purchases, not pilots.

Chroma agrees to join Replit; Chroma Cloud to wind down over the next 60 days

Chroma, the company behind the popular open-source vector database ChromaDB, said it has signed a definitive agreement to join Replit. The combined goal is to build a 'context layer' so agents can draw on everything an organization knows, and Chroma's Foundation memory product will continue inside Replit. Replit's CEO and president were early investors in Chroma. Chroma Cloud will be wound down over the next 60 days, while open-source ChromaDB continues. Early reactions included questions about migration paths for Chroma Cloud users.

Teams running on Chroma Cloud have about 60 days to migrate to self-hosted ChromaDB or another vector store.

Crusoe's Itaipu-powered Paraguay AI cloud draws questions over power contract and Brazil's share

Crusoe announced a partnership with Paraguayan developer Araico to build AI cloud capacity next to the Itaipu dam substation. The first phase is due in mid-2027, with a path to 106MW of IT capacity and options for about 280MW. Regional coverage noted that the US State Department fact sheet gave no tariff, amount or power-utility details. A Paraguayan outlet reported, citing Araico and a utility union, a long-term supply deal starting near $45/MWh. On X, Brazilian commentators debated whether redirecting Paraguay's Itaipu surplus to the project hurts Brazil.

AI capacity is following cheap renewable power to new regions; expect data-residency, latency and geopolitical questions when picking regions.

Fireworks logs a roughly three-hour degradation on GLM 5.3 serving

Fireworks' status page opened and later resolved an incident for degraded performance on its GLM 5.3 endpoint on Oct 23, after a run of similar incidents on DeepSeek V4.1 Flash earlier in the week. Third-party uptime monitors put the episode at about three hours starting around 13:30 UTC, while the main site stayed reachable from all probe regions. Several status bots echoed the incident on Bluesky, and it landed the same day users on X were promoting GLM 5.3 plus DeepSeek Flash combos routed through Fireworks. No root cause was posted.

Repeated per-model degradations argue for multi-provider routing and fallbacks if you depend on Fireworks for open-model inference.

LangGraph patches cross-namespace access bug in Postgres and SQLite stores

LangChain published a medium-severity advisory for LangGraph's Postgres and SQLite stores, where a namespace label containing a dot could resolve to another tenant's namespace. Only the put method validated labels, so get, search, delete, list and batch calls could read, overwrite or delete items belonging to a different namespace. The fix ships in langgraph-checkpoint-postgres 3.1.3 and langgraph-checkpoint-sqlite 3.1.2, and the issue was assigned a CVE and mirrored to the GitHub Advisory Database. It follows a separate advisory this week about the LangGraph CLI printing a registry push token in verbose deploy output, and vulnerability bots on Bluesky amplified both.

Teams using LangGraph stores for multi-tenant agent memory should upgrade the checkpoint packages and audit namespace inputs now.

Inc. profiles why Gamma's CEO rebuilt the product after passing 100 million users

An Inc. column looks at why Gamma CEO Grant Lee chose to rebuild the product from the ground up even though it had passed 100 million users and $100 million in annual recurring revenue. Lee said the guardrails on margins, spacing and fonts that once made decks look polished had become limits as AI improved, and users needed more expressiveness. The rebuild shipped as Gamma 5 on October 6, adding wider visual styles, a freeform canvas alongside agent editing, enterprise connectors and design-preserving import and export. The column presents it as a response to AI output that increasingly all looks the same.

Rebuilding while ahead to fix generic-looking AI output is a playbook for teams whose AI features are starting to feel interchangeable.

Coverage spreads of Perplexity Computer running fully on-device on Macs

Perplexity's new option to run its Computer agent entirely on Apple silicon Macs drew a second day of wide coverage from Mac and tech outlets worldwide. The company says the model, planning and file work all stay on the machine, with Qwen 3.6 or its own post-trained PPLX Pro as the local models and cloud help only when the user approves it. Reporters at MacRumors and Engadget stressed the steep hardware bar of roughly 36 GB of memory and the need for a Pro, Max or Enterprise plan. Reaction on X and Bluesky was largely upbeat about privacy and saving credits, while some commenters questioned the RAM requirement, the subscription cost and Perplexity's long-term prospects.

Hugging Face releases Carbon-X, an open 14B model you can chat with about DNA

Hugging Face released Carbon-X, an open 14.8B-parameter model that reads DNA sequences and answers questions about them in plain English, under an Apache 2.0 license. Built from Qwen3-14B with researchers in Italy and China, it was trained on about 300 billion tokens mixing DNA and text and fits on a single GPU so sensitive sequences can stay local. Hugging Face says it averages 66.7% on five DNA benchmarks, close to a leading frontier model at roughly 1/1,800 of the cost per problem. The team cautions it has not been clinically or wet-lab validated, and the launch drew wide coverage and posts in many languages.

Vercel AI Gateway adds per-session cost and cache views for coding agents

Vercel added a Sessions view to AI Gateway logs that groups requests by conversation and shows the coding agent behind each one, such as Claude Code, Codex, or Cursor. Each session lists its cost, models, providers, and how much input came from cache across turns. Session affinity keeps each turn on the provider holding the prompt cache, which posters said can cut input costs sharply on long agent runs. The CEO promoted it as a unified view of agent sessions and their metadata, and it was the most-engaged Vercel post of the day. Alongside it, posts circulated Gateway data on open-weight models' growing share of tokens and new models arriving on the platform.

Glean lets developers build agents as code from Claude Code, Cursor and Codex

Glean made headless agent building generally available, letting developers create, test and publish Glean agents from coding assistants like Claude Code, Cursor and Codex via a plugin command. Agents are defined as spec and instruction files that live in Git, so teams can review changes in pull requests, run evaluations and roll back bad edits. Published agents still run on Glean's managed runtime with its knowledge graph, permissions across 275-plus enterprise sources and audit trail; only auto-mode agents are supported for now. Reaction on X was positive, with English and Chinese-language explainer threads framing it as agents becoming ordinary software, and trade outlet Unite.AI covered the release.

Profound launches Dynamic Bot Rendering so AI crawlers see fully loaded pages

Profound introduced Dynamic Bot Rendering, which detects visits from AI assistants and serves them a fully rendered copy of a page while human visitors see the normal site. The company argues most AI bots only read raw HTML and miss content hidden behind scripts, tabs or accordions, such as pricing details. A companion Bot View shows what bots see, and customers can choose which pages and which assistants get the rendered version without editing their CMS. It is live for customers on CloudFront, Cloudflare and Vercel. Replies on X were mostly enthusiastic, with one user asking for expiry dates on rendered sale pages.

Start this Monday.