Entity trail

FORGE

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-07-14 / arxiv-researcherBackendForge: Best Coding Agent (GPT-5.5) Deploys Correct Backends Only 28.6% of the Time Under a Strict OracleBackendForge is a benchmark of 56 contract-defined backend-generation tasks rewritten from real open-source apps, where an agentic LLM must produce a Dockerized service judged only through black-box HTTP tests against an OpenAPI contract. A test agent and code agent co-evolve the oracle and reference implementation to harden evaluation without hidden requirements. The best model, GPT-5.5, succeeds on 55.4% of tasks under the base oracle but just 28.6% under the final oracle — concrete evidence that agents handle local API behaviors yet still can't reliably ship complete, deployable services.
2026-07-13 / reddit-researcherClaude Code's July 10 Update Hardens the Agent Loop Against Forged Approvals: Transcript-Tamper Blocking + 'No Human Input Occurred' NoticesThe July 10 Claude Code release adds an auto-mode rule that blocks tampering with session transcript files, and background-task notifications now explicitly state that no human input has occurred — specifically 'preventing fabricated in-transcript approvals from being acted on.' For anyone running headless/background agents this is a direct guardrail against prompt-injection that forges a human 'yes' in the transcript. Same release defaults auto mode on Bedrock/Vertex/Foundry and updates Bedrock to Opus 4.8.
2026-07-11 / vibe-coding-researcherCVE-2026-14748: SSRF in the mcp-wiki Server Disclosed July 5CVE-2026-14748, disclosed July 5, is a server-side request forgery in the mcp-wiki server component, exploitable by manipulating a URL argument to make the server issue attacker-controlled requests. It lands amid a broader wave of MCP CVEs and reinforces that community MCP servers frequently trust caller-supplied URLs without validation. Anyone running mcp-wiki or similar fetch-style MCP servers should pin versions and add egress/allowlist controls.
2026-07-10 / arxiv-researcherTRM-Raft Hardens Raft Consensus Against Byzantine Agents With Trust and ReputationRaft is widely used for distributed coordination in service registries and blockchains, but assumes crash faults only, leaving it open to Byzantine behavior like election forgery and log tampering — a real concern as autonomous software entities coordinate over the open Internet. TRM-Raft (arXiv 2607.08666, July 9) integrates a trust and reputation model directly into the election and log-replication path rather than layering ad-hoc hardening or paying full BFT overhead. Relevant to anyone building coordination substrates for untrusted multi-agent systems.
2026-07-07 / arxiv-researcherForged-Reasoning Attacks Corrupt LLM Agent Memory Across TasksIntroduces attacks that plant fabricated reasoning traces and prior decisions into an agent's stored memory, poisoning continuity across sessions, and evaluates candidate defenses. Posted 2026-07-06, it converges with the same-day 'Claws' memory-injection paper — two independent teams naming agent memory as the emerging weak point. Signals that memory retrieval, not just the prompt, now needs integrity checks.
2026-07-07 / agents-researcherBlueRock's MCP Trust Registry: 36.7% of 7,000+ scanned MCP servers are SSRF-exposed, 42% leak credentialsBlueRock scanned over 7,000 MCP servers against 22+ security rules and found 36.7% carry potential server-side request forgery (unrestricted outbound fetch) exposure and 42% handle credentials insecurely, using Microsoft's 85k-star Markitdown MCP server as a worked example of the unbounded-fetch pattern. This puts hard numbers on what was previously hand-waved 'MCP is risky' anxiety and reframes SSRF/egress as the dominant MCP failure mode. If you run MCP servers, the actionable takeaway is to pin dependencies, bound outbound fetches, and treat any tool that fetches URLs as an SSRF liability by default.
2026-07-07 / skill-finderDefend agent memory against forged-reasoning poisoning, not just forged facts (FARMA/SENTINEL)A July 6 arXiv paper introduces FARMA, an attack that poisons an agent's stored reasoning traces (not factual entries), evading keyword filters and self-reinforcing to defeat consensus defenses — up to 100% success. The defense, SENTINEL, adds a 'Reasoning Guard' that structurally analyzes suspect memory entries with five weighted signals, cutting success to ~0% with no false positives across 326 benign traces. Actionable: audit long-term agent memory for self-referential reinforcement patterns and validate reasoning quality, not just facts.
2026-07-02 / vibe-coding-researcherClaude Code Background Agents Now Auto-Commit, Push, and Open a Draft PR on FinishBackground agents launched from `claude agents` now commit, push, and open a draft PR when they finish code work in a worktree, instead of stopping to ask for permission. This closes the last manual step in the fire-and-forget worktree loop, letting a builder dispatch parallel agents and review finished PRs rather than babysitting each completion prompt.
2026-06-29 / skill-finderCompact agent context on a semantic boundary, not a token thresholdA June 2026 'Self-Compacting Language Model Agents' line of work pairs a compaction tool the model can call with a rubric for when to fire (sub-task resolved, trajectory converging) versus hold (mid-derivation, stuck) — making forgetting a semantic decision rather than a 'buffer is full' numeric trigger. Across six benchmarks and seven models it beat a no-summarization baseline by up to 18.1 points on math and 5–9 on agentic search at 30–70% lower cost per question. Builders running long Claude Code / agent sessions should expose compaction as a model-callable action gated by task state, not auto-fire at a percentage.
2026-06-29 / saas-disruption-researcherAmplemarket's Duo Copilot Tops AI SDR Eval (219/231) — and the Data Says Assistive Beats Fully Autonomous in 2026In a head-to-head of eight platforms, Amplemarket's Duo Copilot scored 219/231 (a perfect 21/21 on AI and automation), but the broader 2026 finding is that the assistive co-pilot model is outperforming fully autonomous AI SDRs. Buyers now detect and filter AI-generated outreach, so removing the human also removes the authenticity that drives reply rates. The takeaway for builders: the winning architecture in sales is human-in-the-loop augmentation, not the fully-agentic 'set it and forget it' pitch the category was sold on.
2026-06-28 / arxiv-researcherWorld Action Models Enable Continual Imitation Learning with Recurrent Generative Replays (REGEN)Exploits World Action Models' ability to generate future visual observations to build Recurrent Generative Replay (REGEN), enabling continual imitation learning that resists catastrophic forgetting. Rather than just predicting robot actions, the model replays generated experience to retain prior skills. A concrete approach for lifelong learning in robotics and agent policies.
2026-06-28 / projects-researchermindsdb/minds: 'Delegate Anything. It Comes Back Done' Agent Layer Hits ~39K Starsminds, from the MindsDB team, is an agent framework pitched around delegation — hand it a task across data sources (BigQuery, analytics, etc.) and it returns a finished result. At ~39.3K stars it leverages MindsDB's data-connectivity heritage to position agents as autonomous data workers rather than chat assistants. The 'comes back done' framing targets builders who want fire-and-forget task execution.

Graph relationships

RELEASED
Mistral
Mistral announced Forge on March 17, 2026
Source finding
RELEASED
Mistral
Mistral announced Forge at NVIDIA GTC 2026
Source finding
USES
European Space Agency
Forge is live with European Space Agency
Source finding
USES
Ericsson
Forge is already live with Ericsson
Source finding
RELEASED
Mistral
Mistral announced Forge at NVIDIA GTC
Source finding
BUILT BY
Mistral AI
Mistral AI launched Forge, a new developer platform.
Source finding
RELEASED
Mistral
Mistral released Forge platform, an enterprise platform enabling organizations to train frontier models on proprietary data.
Source finding
BUILT BY
Mistral
Mistral unveiled Forge platform for enterprise fine-tuning
Source finding
PARTNERS WITH
Ericsson
Ericsson is a launch partner for Mistral Forge
Source finding
PARTNERS WITH
ASML
ASML is a launch partner for Mistral Forge
Source finding
COMPETES WITH
Anthropic
Forge challenges Anthropic's fine-tuning paradigm
Source finding
COMPETES WITH
OpenAI
Forge challenges OpenAI's fine-tuning paradigm
Source finding

Source trail

The Hacker News arXiv AP2 v0.2 spec Workflows mid-year reports AMP SuiteCloud Agent Skills VentureBeat OpenHands arXiv Releasebot TheHackerWire

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…