Interfaze released interfaze-1-lite, an Apache 2.0 open-weight model for OCR and speech that returns confidence scores and bounding boxes with its answers. Teams can send doubtful rows to a reviewer once they have tested how the scores behave on their own documents.
Reality
- Evidence40
- Adoption
- Insufficient
- Hype gap+15
- Incentives70
- Confidence45
Risotto launched an agent for ambiguous Tier-2 IT tickets that saves each new fix as a runbook, and says its customers auto-solve 40% to 90% of tickets. Its bet is that the knowledge from a hard fix, usually lost in a closed ticket, can be kept and run again.
Reality
- Evidence25
- Adoption30
- Hype gap+45
- Incentives70
- Confidence35
Sitefire's SlopShape spots AI-written company blog posts at 97.0% macro-F1 from page structure alone, and at 96.1% after each model rewords its own output. Paraphrasing generated copy hides little from it, though the test reworded text without reordering any page's sections.
Reality
- Evidence40
- Adoption
- Insufficient
- Hype gap+25
- Incentives45
- Confidence35
Omnara has replaced its remote-coding app with an Apache 2.0 control plane that keeps each agent's identity and history while worker machines come and go. Teams get one more self-hostable backend for agents run as durable services, though the only usage figures so far describe the retired app.
Reality
- Evidence45
- Adoption
- Insufficient
- Hype gap+5
- Incentives60
- Confidence50
Cactus Compute released Whistle, an open 16.9MB speech-to-text model that runs on a CPU in the same Needle engine as its tool-calling language models. The accuracy and speed figures are the company's own, so they need rerunning on real devices and audio.
Reality
- Evidence40
- Adoption
- Insufficient
- Hype gap+10
- Incentives60
- Confidence45
Halluminate raised a $30 million Series A led by Oak HC/FT to train AI agents on deal diligence, where the best of seven frontier models averaged 51%. Its customers are AI labs, so the round is a bet that model makers keep paying a nine-person firm for finance-specific training.
Perspective Coverage
4 publishers
- Builder
- Builder 34%
- Operator
- Operator 10%
- Investor
- Investor 56%
Reality
- Evidence45
- Adoption55
- Hype gap+20
- Incentives70
- Confidence55
Jeeves raised $110 million after its stablecoin-settled payments grew from effectively zero to a $1.5 billion annual rate in eight months. The round pays for more countries and more products for companies moving money through markets where banking is expensive.
Perspective Coverage
3 publishers
- Builder
- Builder 23%
- Operator
- Operator 37%
- Investor
- Investor 40%
Reality
- Evidence45
- Adoption55
- Hype gap+15
- Incentives70
- Confidence55
Notion bought ZeroEntropy after its reranker cut latency in Notion's reranking step by 85%, by Notion's count. The hosted product is being sunset with an open-weight release, so teams that built on ZeroEntropy's API now have to decide where the model runs.
Reality
- Evidence40
- Adoption40
- Hype gap+15
- Incentives60
- Confidence45
David AI's 21-language benchmark found speech systems' error rates spread an average of 25.4 points in five Indic languages, against 3.7 in the tightest five. In those languages, choosing a system is a 25-point decision that an English score can make look like a close call.
Reality
- Evidence45
- Adoption
- Insufficient
- Hype gap+10
- Incentives60
- Confidence50
LiteLLM launched Lens on September 30, a tool that uses AI agents to find recurring failures across agent traces sent through its model gateway. Customers host the analyzer and its databases, and the 200,000-trace volume CTO Ishaan Jaffer cites is a future target Lens has not been measured against.
Reality
- Evidence40
- Adoption
- Insufficient
- Hype gap+25
- Incentives60
- Confidence40
Andon Labs opened Pion, its platform for agent-run companies, as a research preview on September 14, with its own agent-run store and cafe still losing money. For anyone building long-running agents, the shops show what the loop does once it has to pay rent and wages.
Reality
- Evidence35
- Adoption
- Insufficient
- Hype gap+45
- Incentives60
- Confidence40
Corgi CEO Nico Laqua told staff in writing that buying pee pads for the office puppy was grounds for firing on the spot, at an insurer one source values at $5 billion. Some Corgi policyholders have no state payout guarantee, so how he gives orders is a matter for the board.
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+20
- Incentives45
- Confidence55
Confido raised a $55 million Series B led by Insight Partners for AI software that automates deductions and trade spend for consumer packaged goods brands. Its chief executive says most customers end up buying the whole suite, and the case for the round depends on that bundle.
Publishers:alleywatch.com · thesaasnews.com · ventureburn.com Perspective Coverage
3 publishers
- Builder
- Builder 24%
- Operator
- Operator 28%
- Investor
- Investor 48%
Reality
- Evidence45
- Adoption45
- Hype gap+30
- Incentives70
- Confidence55
Numeral raised a $100 million Series C led by Insight Partners, almost three times the Series B it closed a year earlier. California's SaaS tax from 2027 adds buyers to a category that Avalara, Vertex, Anrok and Zamp already sell into.
Perspective Coverage
4 publishers
- Builder
- Builder 24%
- Operator
- Operator 32%
- Investor
- Investor 44%
Reality
- Evidence62
- Adoption45
- Hype gap+15
- Incentives70
- Confidence60
The San Francisco startup wants drone fleets to thin clouds over solar farms. The uplift range comes from its own model, the first large flight is targeted for 2027, and standard FAA rules forbid the profile.
Reality
- Evidence22
- Adoption2
- Hype gap+70
- Incentives65
- Confidence55
Nearly ninety percent of the money this company has ever raised arrived in one round. What the syndicate got to underwrite was a customer list rather than a disclosed revenue line.
Reality
- Evidence45
- Adoption40
- Hype gap+20
- Incentives70
- Confidence55
Forbes puts the 18-month-old training-data startup at $3.2bn against the $300m it carried five months ago. Most of that move is multiple rather than growth, and the round has not yet closed or named a lead.
Reality
- Evidence35
- Adoption55
- Hype gap+40
- Incentives65
- Confidence45
Y Combinator's chief said he would do nothing about model distillation, two days after the NSA, FBI and CISA accused six Chinese firms of doing it at scale. Every defence the agencies recommend is the providers' own work.
Publishers:defenseone.com · implicator.ai Reality
- Evidence62
- Adoption
- Insufficient
- Hype gap+15
- Incentives70
- Confidence58
Paper Instruments says agents on its Paper Office libraries and guidance passed 92.5% of trials across 61 Office tasks, against 80.7% on upstream libraries. The test runs the code and the instructions together, so teams have to check on their own files which one earned the gain.
Reality
- Evidence35
- Adoption
- Insufficient
- Hype gap+25
- Incentives65
- Confidence40
Specific Labs scores coding agents on licensed production codebases. The best setup clears 38.8%. The analysis covers ten tasks at eight runs each, so every published score is a count of passing rollouts out of 80.
Perspective Coverage
3 publishers
- Builder
- Builder 52%
- Operator
- Operator 33%
- Investor
- Investor 15%
Reality
- Evidence48
- Adoption
- Insufficient
- Hype gap+30
- Incentives70
- Confidence58
Earlier coverage
- Balderton takes up to 90% of Spott's $21 million Series A
Invest · September 22, 2026 · 4 publishers
- Tailscale doubles its business customers to 40,000 since last November
Invest · September 23, 2026 · 1 publisher
- Firecrawl raises $75M, launches Alexandria to cut custom connector work for AI agents
Product · September 22, 2026 · 1 publisher
- Nearly half of one newsletter's 25 product openings ask for eval experience
Invest · September 22, 2026 · 1 publisher
- Dilution alone cuts a 0.25% startup grant to 0.153% by Series C
Invest · September 22, 2026 · 1 publisher
- Blacksmith enforces its org-only rule by leaving the job queued forever
Build · September 20, 2026 · 1 publisher
- Perseus fires a $2,500 kinetic interceptor at Yuma under Falcon Peak 26.2
Product · September 19, 2026 · 1 publisher
- Dipole Labs aims optical circuit switches at the dead time GPUs spend waiting on the network
Build · September 19, 2026 · 1 publisher
- OS3 prices hotel towel folding at $2.50 an hour on hardware it built down to the actuators
Build · September 18, 2026 · 1 publisher
- Cua trains a 2.8 MB scorer to replace the per-field reasoning step in form agents
Build · September 18, 2026 · 1 publisher
- Two thirds of Angle Health's $600m round buys out early holders at $2.5bn
Product · September 18, 2026 · 2 publishers
- Raindrop replays real production traffic against every pull request to test agent changes
Product · September 18, 2026 · 1 publisher
- Apollo's Watcher escalates a flagged agent action to a bigger AI before any human sees it
Product · September 17, 2026 · 1 publisher
- Raindrop replays production agent traces against synthetic services on every pull request
Build · September 17, 2026 · 1 publisher
- Flock's nationwide lookup returned Boston plate reads for three days after the city confirmed sharing was off
Build · September 15, 2026 · 1 publisher
- Seven buyers put $82.06m of preferred stock into Firecrawl at 5.7 times its Series A
Build · September 14, 2026 · 1 publisher
- Nomic's CEO alleges startups resell frontier traces generated on lab credits
Build · September 14, 2026 · 1 publisher
- Andon Labs' Pion hands one agent the email, phone, banking and cards of a whole business
Build · September 14, 2026 · 1 publisher
- Superhuman agrees to buy Fathom after testing its own notetaker on selected users
Build · September 14, 2026 · 1 publisher
- Two rounds account for all of Nvidia's $1.3 billion in August lead deals
Invest · September 14, 2026 · 1 publisher
- Atomarine plans a gas-fired floating data center, with its first reactor four years later
Product · September 13, 2026 · 1 publisher
- Orchestra builds its cheaper-model switch gate from the buyer's own corrected traces
Build · September 13, 2026 · 1 publisher
- Investors are pricing a head start that an acquirer's legal team can delete
Invest · September 12, 2026 · 1 publisher
- Meta's AI chief credits evaluation systems, not raw intelligence, for agents beating 100 engineers
Invest · September 12, 2026 · 1 publisher
- Garry Tan argues American open-weight labs should distill American frontier models
Product · September 11, 2026 · 1 publisher
- Corgi added twelve dollars of valuation for every dollar of its $106m round
Invest · September 11, 2026 · 1 publisher
- Garry Tan relocates AI's moat from the model weights to the price list
Invest · September 10, 2026 · 1 publisher
- Baseten buys Blaxel to put the agent sandbox on the same network side as the model call
Build · September 10, 2026 · 1 publisher
- Naming the object was worth 0.150 to SAM 3 before Feyn Labs trained anything
Build · September 10, 2026 · 1 publisher
- Embedder hands its firmware agent flash-and-halt authority over SEGGER's J-Link probes
Build · September 10, 2026 · 1 publisher
- Fine-tuning hands the open-weight safety question to the buyer
Leadership · September 3, 2026 · 1 publisher
- Conveo's $50M round doubles as a buyer's checklist for AI-moderated research
Leadership · September 3, 2026 · 1 publisher
- HFlow separates the black-frame measurement from the threshold that rejects the episode
Build · August 31, 2026 · 1 publisher
- An unsupervised agent loop billed $38 before anything in the system said stop
Build · August 30, 2026 · 1 publisher
- CoArena licenses the preference labels its free leaderboard generates
Build · August 29, 2026 · 1 publisher
- A 90-pound panel every 30 to 40 seconds is what finances Cosmic Robotics' general-purpose platform
Build · August 29, 2026 · 1 publisher
- Several dozen Alpharetta cameras write permissions for 2,000 outside organizations
Build · August 29, 2026 · 1 publisher
- Experiential Labs bets its open-source router's traces will train cheaper replacements for rented models
Build · August 27, 2026 · 1 publisher
- Updating 5% of rows drove one BRIN query from 1,536 heap pages to 51,268
Build · August 27, 2026 · 1 publisher
- Ox Alpha passes the Xinjiang test and fails on Xi: seven topics, 83 points apart
Build · August 25, 2026 · 1 publisher