Chamelio raised a $26M Series A led by Entrée Capital, just five months after its $10M seed round in January 2026, with ARR up 4x over that stretch and total funding now at $36M. The signal: in-house legal teams are starting to hand off contract review, negotiation, and execution — work that used to require a lawyer's line-by-line judgment — to AI agents trained to act on legal work autonomously.
Ema raised a $77M Series B led by Bengaluru-based Creaegis, with existing investors Accel, Section 32, and Prosus increasing their stakes, bringing total funding to $140M at more than 4x its 2024 valuation (amount undisclosed). The signal: Ema's 'AI employees' are taking over work enterprises used to outsource to SaaS products and IT services firms — AI is now competing directly for that same budget line.
Firecrawl raised a $75M Series B led by Smash Capital, exactly one year after its $20.7M Series A, bringing total funding to more than $95M. The real signal isn't the amount — it's the same-day launch of Alexandria, a platform that pays researchers, developers, and public institutions for their knowledge and resells access to AI agents, marking Firecrawl's shift from 'web-scraping tool' to 'knowledge supply chain for AI agents.'
Claude Opus 5.5 shipped 2026-09-22, model ID `claude-opus-5-5`; 1,000,000-token context window (128K output cap); API pricing is $4.00 input / $20.00 output per 1M tokens (down from Opus 5's $5/$25), with cache reads cut 60% to $0.20; Terminal-Bench 4.0 jumps from Opus 5's 52.3% to 66.4%, and GDPval-AA v2.1 leads at 1846 Elo versus Fable 5.1's 1735 and GPT-6 Astra's 1542; Anthropic's own numbers show it still trailing GPT-6 Astra on AutomationBench and Terminal-Bench-Science; because its biology and cybersecurity capabilities now match Claude Mythos 5.1, it ships with Fable-5.1-level safeguards, and on the API side thinking mode can no longer be turned off while forced tool use now returns an error instead of failing silently
GPT-6 Sol (`gpt-6-sol`) / GPT-6 Luna (`gpt-6-luna`) shipped 2026-09-22; 1,050,000-token context window (922,000-token max input, 128,000-token output cap — same as flagship Astra); Sol prices at $2.00 input / $10.00 output, Luna at $0.10 input / $0.50 output per 1M tokens (a further 50% cut versus GPT-5.6 promotional pricing); Sol hits 33.2% on AutomationBench at xhigh effort, beating Claude Opus 5's 26.9% at roughly 1/11 the cost; DeepSWE v1.1 puts Sol at 68.8% and Luna at 66.6%, closing in on Claude Fable 5.1's 69.9%; in OpenAI's internal simulated deployment testing, severity-3+ misalignment flags dropped from 66 to 42 versus the prior generation
The most common trap in growth interviews is describing a 'viral loop' as one line — 'let users invite friends' — without answering whether the loop feels like spam to the person on the other end, or what metric decides whether it's worth keeping. Today's practice question: a B2B project management SaaS has stalled, and the CEO wants a growth engine that doesn't rely on paid ads, while users already create tasks and export weekly reports to clients. The framework is the Hook Model (Trigger–Action–Variable Reward–Investment) to check whether an existing repeated behavior can naturally become a touchpoint that reaches outsiders, paired with growth-loop thinking to turn a one-shot funnel into a loop that resupplies itself. The case study is Fyxer — a four-person growth engineering team that ran 541 experiments in a year at a 25% win rate, testing which of users' already-repeated actions could become a growth loop, and took ARR from $1M to $35M — while also killing a loop that sounded clever on paper but felt like noise to users.
The Gates Foundation released its 2026 Goalkeepers Report on Sept 21, pledging at least $1B over two years for equitable local-language AI, alongside a joint pledge from 60 organizations — including Anthropic, Google, Microsoft, and NVIDIA — to reach 3.4 billion people with local-language AI over five years; yet in the same 72-page report, Nigeria — Africa's most populous country and one of its most active AI hubs — gets only two footnote mentions, while Kenya, Sierra Leone, and Rwanda each receive full guest essays; a Kenyan tech outlet argues the same week that the country's push toward agentic government services is stuck on an old problem — government systems still lack interoperable APIs and governance frameworks; and Norrsken22 partner Lexi Novitske observes that as US private capital retreats from African tech, the gap is being filled fast by Chinese open-source models like Qwen and DeepSeek, echoing the earlier playbooks of Transsion and Huawei in hardware, and OPay/PalmPay in fintech.
AI coding startup Cognition (valued at $48B) opened its first Latin American office in São Paulo on Sept 22, after already landing more than ten major Brazilian enterprise clients remotely, including Nubank, Itaú, and Santander; Chile's Chamber of Deputies passed an anti-deepfake bill on first reading, 128 in favor, 2 against, 5 abstentions, with fines of up to 716 million Chilean pesos, though the regulator isn't expected to be operational until December and a proposal would push the effective date back a full year, while Brazil and Paraguay advance similar legislation; and the EU launched the EU-LAC AI supercomputing network in Córdoba, Argentina on Sept 22, a two-year collaboration spanning Argentina, Brazil, Chile, Colombia, Costa Rica, Mexico, and Uruguay.
Attackers stole MemTensor's GitHub Actions publish tokens and pushed malicious releases of the npm package @memtensor/memos-cloud-openclaw-plugin (0.1.21/0.1.23/0.1.25) and the PyPI package MemoryOS (2.0.34). The npm plugin launches a Go-based stealer called sckit when the OpenClaw agent gateway starts and again on every memory-recall event — passing the current user prompt to the malicious binary. The PyPI package triggers on a bare `import memos` via a hooked logging initializer. sckit harvests npm/PyPI/GitHub/GitLab/AWS/Vault/SSH credentials and exfiltrates to a C2 under skyleen[.]fr, with built-in code to re-publish itself into other packages and GitHub Actions workflows. Socket, StepSecurity, SafeDep, and Aikido independently confirmed the compromise. Fix: pin to a clean version, rotate every credential the affected host could reach, and block the C2 domain.
petit-poucet is a Rust-built MCP server that stores AI coding agents' long-term memory as Markdown notes in a git repository. Install: claude plugin marketplace add https://github.com/areguig/petit-poucet, then claude plugin install petit-poucet@petit-poucet. It solves the problem of every agent tool keeping its own separate memory, so switching tools or starting a new session means repeating yourself.
OpenAI shipped GPT-6 Sol/Luna about an hour after Anthropic's Opus 5.5 launch; Grok 4.7 scored 38% on its own Terminal-Bench 4.0 run versus 26% on independent re-testing — frontier labs are now shipping same-day counter-launches. Anthropic, OpenAI, SpaceXAI, and Google were named together in an antitrust suit for the first time. AWS AgentCore, MaxKB (CVSS 10.0), and a MemTensor MemOS supply-chain attack, plus three papers on Loopjacking, APort Vault, and authorization revocation, all converge on one point: review, model judgment, and cancellation are not real verification. Ema, Chamelio, Firecrawl, and Enhans raised over $200M combined in a single week, betting that agents will take over work enterprises used to outsource.
RRSI finds that the strongest existing harness self-improvement method actually finishes 1.7 points below doing nothing when moved to unseen tasks, while adding regularization pushes out-of-distribution performance more than a point above the starting harness; SkillSpec uses Hoare-style specification reasoning to catch 239 out of 515 real-world agent skills with confirmed defects, at 61.2% precision; Jev-Mem offloads high-frequency memory decisions to a non-generative System-One controller, lifting LoCoMo score by 11.0%, construction speed by 6.6x, and cutting query latency by 36.7%
browser-use/video-use lets Claude Code edit video directly, using ElevenLabs transcripts to find cut points, at 25,702 stars; dream-num/univer repositions its office SDK as an 'Office Harness for AI Agents' and tops today's TypeScript trending; superdesigndev/treg is 'OpenRouter for agent tools,' letting agents call 3,000+ metered tool endpoints with no contract required; davila7/claude-code-templates crosses 30K stars by replacing hand-rolled config with one-line agent/command/MCP template installs; pydantic-ai v2.47.0 tightens type validation so a bad UserPromptPart.content type no longer silently degrades.
Grok 4.7 shipped 2026-09-21 as `grok-4.7`: a new, larger base model with longer RL training; 500K-token context window; API pricing unchanged at $2.00 input / $6.00 output per 1M tokens (rising to $4.00/$12.00 above 200K tokens). xAI's own Terminal-Bench 4.0 score jumps from 20.3% to 38.0%, but Artificial Analysis's independent re-test puts it at only 26% — well behind GPT-6 Astra (60%) and Claude Fable 5.1 (55%). It rolled out to all GitHub Copilot plans the same day. For agent builders, it's a 3-8x cheaper option for agentic coding, but complex autonomous terminal work still warrants independent benchmarking before you rely on it.
Lasso Security found that any MaxKB assistant with a tool, MCP tool, skill, or sub-application attached gets a SandboxShellBackend that bundles in a shell execute tool — one MaxKB never excludes from the tool list and never adds to the human-approval list, so it runs with zero oversight. That leaves the shell tool wide open to anything the agent reads: support tickets, RAG-ingested documents. Public or embedded anonymous assistants need no privileges at all to trigger it, earning a perfect CVSS 10.0. Patched in v2.10.5-lts; no evidence of in-the-wild exploitation.
Loopjacking reproduces a 'human approves A, system executes B' failure mode in real released products — Agno AgentOS, LangGraph Agent Server, and OpenClaw; APort Vault runs 225,964 evaluations to show that what actually stops an agent from making unauthorized payments is a deterministic policy layer at the tool boundary, not a smarter model; Authorization Revocation formally proves that 'cancellation' may not even be a well-defined concept once delegation and asynchronous execution are involved
An antitrust suit puts Anthropic, OpenAI, SpaceXAI, and Google in the same case for the first time; the UN's first AI science-panel report warns there's no assurance humans keep control over AI agents; today's Arxiv Digest and a real AWS AgentCore breach both prove that review, model judgment, and a cancel button don't equal verified authorization; Grok 4.7 undercuts on price but trails badly on benchmarks; SoftBank borrows over $11B for its OpenAI stake; Taiwan's sovereign medical-AI push, India's regulatory stance, and Southeast Asia/Africa regional updates round out the day
Microsoft open-sourced agent-governance-toolkit (6,303 stars), enforcing tool-call policy in code instead of prompts, citing an ICLR 2025 paper showing 100% adaptive jailbreak success on GPT-4o/Claude 3/Llama-3; ai-memory grew from 2,900 to 7,575 stars in a month, giving 20+ coding agent CLIs a shared long-term memory; anthropics/financial-services ships the same finance-vertical agents as both a Cowork plugin and a Managed Agents API template, at 35,728 stars; the official MCP Inspector reached v2.7.0, unifying its web/cli/tui clients into one binary; coder/coder folds AI coding agents into Terraform-defined, controlled dev environments with no API keys in the workspace.
Today's Deep Learning & NLP rotation covers five core concepts: how self-attention scores relevance with Query/Key/Value, why positional encoding is a necessary patch for self-attention's permutation-invariance, how subword tokenization (BPE/WordPiece) solves the out-of-vocabulary problem, how to answer the classic fine-tuning-vs-RAG follow-up, and how embeddings represent semantic similarity as vector distance. The practice question comes from a Google agentic AI engineer interview guide: explain how attention works and how context compression techniques affect retrieval accuracy — the breakdown threads attention's computational complexity together with the compression-vs-accuracy trade-off into one complete answer.
The most common trap in metrics interviews is treating 'this number went up' as proof that a decision was correct, without first separating what the north-star metric, driver metrics, and guardrail metrics are each protecting. Today's practice question: an A/B test shows a new checkout flow raised revenue but customer satisfaction (CSAT) dropped — should you ship it? The framework is to sort every number into a metric tree first, then dig into the mechanism behind the CSAT drop (did returns or complaints move too), instead of ruling on a single metric's direction. The case study is Netflix's former CPO Neil Hunt picking exactly one north-star metric — 'grab the viewer's attention within 90 seconds' — for artwork A/B tests, so a single title's engagement or overall watch hours couldn't hijack the final call. Swapping in a cover image of a kid playing golf with his caddy for 'The Short Game' alone lifted click-through by 14%.