labloop Week –
Upskilling plan · v4 · replaces v3 (Vara, 2026-09-10)

Applied AI for Biotech R&D and Biomanufacturing

Twenty weeks at 8–10 h/week, then a six-month commercial phase. Three small open instruments, one priced offer, and a first invoice by week 16. The name is not the point; the number is.

Author
Peter Ciancaglini
Dated
2026-09-11
Horizon
20 weeks + 6 months
Program
labloop
This week–
Next gate–
Founder conversations0
§0 · Read this first

What changed from v3, and why

Vara is not started as a flagship. Two of its ideas, the telemetry schema and the calibrated judge, carry over into the instruments.

Topicv3 · Varav4 · this planOrigin
AudienceDevs, NGOs, Spanish-speaking startups in generalWet-lab biotech, biomanufacturing, agbio and pharma R&D: organizations with money, complex problems and a slow, expensive experimental loopYour ask + the GridX list (64 of 86 startups have no AI/IoT base)
FlagshipOne benchmark (Vara)Three small instruments that each close one gap the list shows repeatedly, plus one priced offerInstruments are your talent; selling them is your trap
TheoryLLM internals (Karpathy, CS336)Bioprocess, DoE, Bayesian optimization, multivariate monitoring, hybrid models, on-prem document AI, plus a science literacy thread"Professionalize in AI science"
MoneyOffer expected in month 6–9First invoice by week 16 (any amount). The charging muscle is trained inside the plan25 years, USD 0 charged
Publishing1 build log/week, 3 deep divesGitHub 3–4 commits/week, LinkedIn 2/week, Medium 4, arXiv/bioRxiv 2, Zenodo DOIs 3, 60-s videos 3Your ask
HardwareCloud APIsRTX 3050 + Debian boxes are a feature: "your lab's AI runs on a USD 300 computer and your data never leaves the building"Your constraint, reframed
Culture goalNot addressedExplicit filter for the next employer or client: scientist-led, biology-paced, no delivery-pressure cultureEY exit

Hours stay at 8–10 h/week. If you can do more, the calendar compresses; it never expands in scope.

§1 · Thesis

A data-to-decision problem, not an AI problem

Most biotech startups do not have an AI problem. They have a data-to-decision problem.

Their experimental loop takes weeks because the data lives in spreadsheets, instrument exports, paper batch records and PDFs. The AI that matters to them is not a chatbot. It is the layer that closes the loop: structured capture, models that work on small and expensive data, and a decision every week instead of every quarter.

I build that layer, measure how many days it takes out of the loop, and publish the number.

design run measure analyze decide next run LOOP CYCLE TIME weeks → days a decision every week
The experimental loop every instrument shortens. Today, each turn takes weeks because the data between steps is not machine-readable.

Brand metric 1

Loop cycle timedays

Days from hypothesis to decision.

Brand metric 2

Data readiness%

Share of experimental data that is machine-readable within 24 h of being generated.

Every instrument moves one of the two. Every offer measures one of the two first.

Why this niche is yours

  • Same talent you showed at EY. You made delivery visible and measurable when nobody asked. The lab equivalent is "how many days from question to decision, and where do those days go". Nobody in a wet lab has that number.
  • The evidence is in the list. 64 of 86 GridX startups have no AI/IoT base. Their gaps cluster into four repeated patterns, and none needs a new algorithm. They need an engineer who understands the science, the software and the money.
  • The money is real. Pharma, CDMOs, agbio groups and funded biotech pay for speed and for risk reduction. Sell weeks saved per experiment cycle, converted into runway. Startups buy time. Pharma buys risk.
  • Biology sets the pace. A fermentation takes 7 days, a field trial a season, a stability study months. That timescale structurally prevents "deliver more with less by Friday".
  • AI-native biotechs are peers, not clients. Tesabio.ai, Zavia, Nanogrow, CellCo, Dogma, Mavios and the other "AI: yes" companies: learn from them, cite them, ask them to review or co-author. Sell to the 64 that are ten years behind in data maturity.
§2 · Strategy

How the dots connect

Theory feeds instruments, instruments produce evidence, evidence opens three audiences with the same artifacts.

THEORYINSTRUMENTSEVIDENCEAUDIENCESOUTCOME Theory Bioprocess, DoE,Bayesian optimization,MSPC, hybrid models,document AI Vara ideastelemetry schema +calibrated judge 1 · batchlensWK 1–8 fermentation batchintelligence 2 · trialkitWK 9–13 trials and assaysto a decision 3 · labdocsWK 14–18 on-prem lab documentintelligence Evidence repos + releases3 Zenodo DOIs2 preprints4 deep dives3 videos1 case study 64 non-AI startupsGridX: feedback, data,co-authors, first paidaudits Established + fundspharma, agbio, CDMOs,funds: high-ticketaudits and pilots Scientific-data cos.TetraScience, DataHow,Aizon: employee track,remote, relaxed Loop Audit2 weeks, fixed price Pilot6–8 weeks Retainerfractional AI lead Job offersfallback or by choice
Dashed edges: carried-over Vara ideas feed batchlens (week 5 harness) and labdocs (week 15 judge); the employee track runs in parallel. You never sell the instrument; you sell the number it produces about their loop.
  • Theory feeds instruments. Every lecture in §6 exists because a practice in §7 needs it. Nothing is studied "for later".
  • Instruments produce evidence, not products. Public data first, then one founder's sample for free, then inside a paid audit.
  • Evidence opens three audiences at once. The startup list is the proving ground and first invoices; established companies and funds are the high ticket; scientific-data companies are the employee track.
  • The offer is fixed-price and small on purpose. A 2-week Loop Audit sits below most purchase thresholds, trains the muscle you have never used, and generates the case study that sells the next one.
  • Funds are the channel. GridX, SF500, CITES each sit on 30–100 portfolio companies with the same four gaps. One portfolio package puts you in front of ten startups without ten sales conversations.
  • The Vara work is not lost. Its telemetry schema (cost × model × task) and calibrated LLM-judge make instruments 1 and 3 measured, not claimed.
§3 · Target audience and companies

The GridX list, read as a 96-well plate

All 86 descriptions re-read and grouped by each company's primary gap. Four archetypes (A–D) are the sellable ones; E waits for phase 2; F are peers.

AProcess & telemetryBField trialsCLab assaysDTraceability & QMSEDiagnostics (phase 2)FAI-native peers·Off-thesis
One well per company, filled in archetype order (rows A–H, columns 1–12); 10 wells stay empty. A heavy outer ring marks the 22 companies that declare an AI/IoT base. Hover a well for the company; click it to open its row in the company map.

Primary fit, 86 companies

AProcess & telemetry25
BField trials8
CLab assays & preclinical14
DTraceability & QMS7
EDiagnostics & clinical11
FAI-native / computational20
·Off-thesis1

A–D together: 54 companies with a sellable gap. AI/IoT base declared: 22 of 86.

Rules for using this list in public

Never name a company's weakness in public. The CSV's gap column is your private hypothesis, not a fact. Post about archetypes ("a fungal pigment fermentation startup"); name companies only to praise them or after they engaged.

CodeArchetype and typical stated gapCos.InstrumentSells as
AProcess and telemetry: bioreactors, photobioreactors, digesters, incubation rooms, extrusion lines, cold chain. Manual monitoring, paper batch records, no golden batch, no early deviation detection
Michroma, Levya, Fungilife, NanoInGreen, POLYMERA, Protiva, Food4You, Bioplastix, BreakPET, Hiamet, Caligenia, ByBug, RADIAL, algaebio+, Algalife, Tomorrow, Einsted, CyanoMin, Alkemio, Enzyva, BioBlends, Nat4Bio, BioPlaster, Neocell, Calfix (weak)
251 · batchlensBatch intelligence audit; soft-sensor pilot; "next run" optimization
BField trials: agro bioinputs, aquaculture. Trials scattered in spreadsheets, subjective visual scoring, no consolidated efficacy report
Puna Bio, Nunatak, Bsafe, Semion, MicroIN, Qumir Nano, AQUIT, FeedVax
82 · trialkit fieldTrial data consolidation + statistically defensible efficacy report
CLab assays and preclinical R&D: microplates, kinetics, dose-response, formulation stability. Excel analysis, slow reports, fragmented lead-selection data
Resistia, APEXzymes, Antarka, Syocin, Arakion, Scitherm, BioClé, Lipock, Kresko, Nanotransfer, Invitrall, Bioeutectics, Vexxel, Werken Vac
142 · trialkit plateAssay pipeline + automated report; formulation data model
DTraceability, QMS, regulated manufacturing: audit trails, batch genealogy, clean-room records. Regulatory traceability demanded and absent
Eywa Biotech, AVaTAR MedTech, NanoproX, Apasomics, Synthetic Ocular Restoration, SphereBio, New Organs Biotech
73 · labdocsOn-prem document intelligence + ALCOA+ audit trail; deviation report drafting
EDiagnostics and clinical reporting: point-of-care readers, microbiome and genetic reports, fertility clinics. Manual reports, no portal, no HIS/LIS integration
CASPR, Gisens, Ubique, ZEV, NanoJump, Enteria, Microgenesis, PreGen, Embryoxite, Fecundis, GlycoX
114 · phase 2Clinical report automation with validation; regulated, higher friction
FAI-native and computational platforms. Gap is software engineering and MLOps maturity, not AI
Tesabio.ai, Nanogrow, CellCo, Dogma, CellRep, Elytron, Zavia, Bitgenia, MIGMA, Naturannova, Mavios, Geoprot, Monte Caldera, Qnity, Beeflow, Biodiversity Intelligence, ArgenTag, OncoPrecision, Stämm, Sylvarum
20noneDo not sell to. Learn from, cite, ask for review, co-author, endorse on arXiv

Tier the 64 non-AI companies by what you need from them

Not by what they can pay. Click a dot when a partner says yes.

Feedback partners

Five founders who will look at an instrument for 30 minutes.

Data partners

Two who share a sample dataset under NDA for a free run.

Co-author or endorser

One founder-scientist (most are CONICET researchers) for co-authorship or arXiv endorsement.

Beyond the list: who actually pays high ticket

Unverified research list

Current status, funding and openings of these companies are not verified. Each name needs 10 minutes of checking before outreach.

CompanyWhyWhat to offer
Bioceres / Rizobacter RosarioAgbio group with bioinputs, field trials at scale, own VC (SF500)Trial data platform audit (B); portfolio package via SF500
Insud Pharma: mAbxience, Elea, ChemoBiosimilars manufacturing, GMP, PAT and data-integrity budgetsBatch intelligence (A) and audit-trail (D) pilots
Biogénesis Bagó, Sinergium Biotech GarínVaccine manufacturing, fermentation and cell culture, regulatedA and D
Gador, Laboratorios Bagó, Roemmers, Richmond, Biosidus, Amega/Gema BiotechPharma manufacturing and biotech in Buenos Aires provinceD (data integrity) and A
Terragene RosarioBiological indicators for sterilization, sells worldwide, lab-heavy QCC and D
Porta Hnos, Bio4, ACA Bio, LedesmaIndustrial fermentation (ethanol) at scale, plants full of sensors and thin on analyticsA
Y-TEC YPFApplied research, lithium and biotech, budgetsA and C
Mastellone, Quilmes/AB InBev, PeñaflorDairy, brewing, wine: fermentation with QC labsA (yeast, fermentation curves)
Regional: ATGen, Genlives (UY) · Biomm, Eurofarma, Aché, Cristália, Bio-Manguinhos, Butantan, Raízen (BR) · Andes Biotechnologies (CL) · Landsteiner, Sophia, Probiomed (MX)Regional expansion once case studies exist (month 6+)Same offers, in Spanish and Portuguese

Funds and programs as channels

GridX (contact)SF500CITESDraper CygnusKamay Ventures

Offer: "portfolio data-readiness package", 5 startups × 1-week mini audit, one consolidated report for the fund's partners.

Scientific-data and bioprocess-AI companies

TetraScienceUncountableBenchlingScispotSapioLabKeySynthaceDataHowNovasignBasetwoAizonInvertQuartic.aiCulture BioGinkgoCradleLatchBioSartorius/UmetricsCytivaNovonesisDSM-FirmenichLonzaNovo NordiskNNE

Inspiration now, employer track later. They will read your preprints. batchlens is, on purpose, a small open version of what DataHow and Aizon sell.

Not on the list on purpose: Big Four and large IT consultancies' life-sciences practices. Same culture you are leaving, different logo.

Culture filter: ask before signing anything

  1. Who decides priorities: scientists, or a delivery office?
  2. What happened the last time a milestone slipped? Listen for "we re-planned" versus "we worked weekends".
  3. How long is one experimental cycle here? If the answer is in days, the company is paced by biology. Good.
  4. Is there on-call, and for what?
  5. How many people left in the last 12 months, and why?
  6. Can I see the last two retrospectives or lab meeting notes?

Two bad answers out of six: walk away. Calibration point: ask your friend at Sumble to answer these six about their own workplace so you know what "good" sounds like.

§4 · Offers and pricing hypotheses

Four offers, fixed price, half upfront

These prices are hypotheses to test with the first five conversations, not facts.

One-off engagements, USD

Loop Audit seed, case-study price
2,500
Loop Audit Series A+, established, CDMO
6,000–10,000
Portfolio package 5 mini audits for a fund
12,000–18,000
Pilot 6–8 weeks, one instrument
12,000–30,000
05k10k15k20k25k30k
The retainer is recurring and sits off this axis: USD 4,000–8,000 per month, 1–2 days a week.
OfferWhat it isDurationWho buys
Loop AuditMap the experimental loop, inventory data sources, measure cycle time and data readiness, run one instrument on their sample data, deliver 3 automation candidates with estimated days saved and one working prototype2 weeks, ~20 hFounders, Head of R&D, QA managers
Portfolio package5 mini audits + one consolidated report for the fund5 weeksFunds and programs
PilotOne instrument adapted to their data and running in their lab, with a measured before/after6–8 weeksAnyone who did an audit
Fractional AI lead for R&D1–2 days/week, owns the data-to-decision layerMonthly retainerCompanies with 2+ instruments running

Pitch in runway, never in savings

"One fewer week per cycle × N cycles a year = M months of runway you did not have to raise." For pharma: "batches not lost, findings not written".

week(s) saved × cycles/yr = 2.8 months

Terms

Fixed price, half upfront, paid by invoice or card. No hourly billing in the first year: hourly reproduces EY's logic inside your own business.

§5 · The three instruments

Seven steps that cure the instrument trap

Every instrument goes through the same sequence. Step 7 is the one that matters most.

  1. Build on public data.
  2. Publish the number it produces: deep dive, DOI, preprint.
  3. Run it free on one founder's sample: 30 min, NDA if needed.
  4. Publish the anonymized before/after.
  5. Put it inside a priced Loop Audit.
  6. Ask the client for a quote you can publish.
  7. Move on. Never add a feature no client asked for.

batchlensfermentation batch intelligence

weeks 1–8A archetype Apreprint 1 · wk 8

Gap it closes

Paper or spreadsheet batch records, no golden batch, deviations detected at harvest, yield optimized by intuition.

What it does

ingest batch CSVs or paper recordsbatch unfolding + alignmentgolden-batch envelopePCA, Hotelling T², SPEsoft sensor: biomass, titerearly deviation alarmsBO next-run feed profileHTML report per batch

--fake mode runs everything with no data.

Public data

IndPenSim: industrial-scale penicillin fermentation simulator, 100 batches with online process variables and Raman spectra, including faulty batches (verify the link on industrialpenicillinsimulation.com). Plus your own Monod fed-batch simulator streamed over MQTT, so the IoT gap is demonstrable without hardware.

Stack

pandasscikit-learnLightGBMstatsmodelsBoTorch (CPU)PlotlyMosquittoNode-REDInfluxDBGrafanaESP32 + probe, USD 15

Headline numbers you publish

  • Hours-to-detect a deviation versus harvest time
  • Titer prediction error at 50% of batch time
  • Simulated titer gain after 12 BO-suggested runs versus a DoE grid

Hardware fit

All models train on CPU or in seconds on the RTX 3050. The 6 GB card matters only in week 5 (vision model for batch-record extraction).

Show it to

Michroma, Levya, Fungilife, NanoInGreen, POLYMERA, Hiamet, Food4You. Peers for review: Mavios (they sell the enterprise version), Stämm.

Preprint 1

"Early deviation detection and soft sensing on an open industrial fermentation benchmark with consumer hardware: a reproducible baseline". arXiv cs.LG or q-bio.QM (needs endorser), or bioRxiv. Zenodo DOI for the code.

trialkitfrom trials and assays to a defensible decision

weeks 9–13B fieldC plateoffer launch · wk 13

Gap it closes

Field trials and plate assays scattered across spreadsheets, subjective scoring, hand-built reports, no confidence intervals, so investors and growers do not trust the efficacy claim.

What it does

LLM maps any spreadsheet to one trial schemadesign detection: RCBD, split-plotmixed model, effects + intervalsplate: logistic/Gompertz, 4PL, MIC, Michaelis-Menten, Z′image scoring: CLIP → small classifierreport in Spanish or English

Public data

Field: agridat (hundreds of published trials via the Rdatasets mirror), Genomes to Fields. Plate: published growth-curve and dose-response sets, PlantVillage for image scoring. Zindi/CGIAR challenges for practice.

Stack

statsmodels MixedLMscipy curve_fitPyTorch small CNNCLIP on the 3050Claude or local QwenVara harness

Headline numbers

  • Days from last measurement to report, manual versus trialkit
  • Mapping accuracy of spreadsheets to schema
  • Agreement between image scoring and human scoring

Show it to

Field: Puna Bio, Nunatak, Bsafe, Semion. Plate: Resistia, APEXzymes, Antarka, Syocin. Peers: Beeflow, Zavia.

Output, week 13

Medium deep dive, Zenodo DOI, 60-s video, and the Loop Audit offer launched.

labdocson-prem document intelligence with an audit trail

weeks 14–18D archetype Dpreprint 2 · wk 18

Gap it closes

SOPs, protocols, batch records and notebook pages that are not searchable, traceable or auditable, while ANMAT, FDA and EMA demand ALCOA+ traceability the startup does not have.

What it does

PDF/scan → local VLM → structured experimenteLabFTW via API, hash-chained audit traillocal search + Q&A with citationsdeviation drafts from batchlens alarms + SOPcalibrated eval: 60 labeled cases

Public data

Synthetic batch records and SOPs generated from real templates (never client data), plus public regulatory documents. eLabFTW on the Debian box as substrate: "AI inside a real ELN", not a toy.

Stack

Ollama / llama.cppQwen2.5 7B Q4Qwen2.5-VL 3BGemma 3 4Bbge-m3 / nomiceLabFTW + DockerPostgresVara judge

Headline numbers

  • Extraction accuracy per field, local versus cloud
  • Judge-human agreement
  • Minutes from scanned page to auditable record

Hardware fit

Where the small boxes become the argument: what runs on a USD 300 mini PC, what needs the 3050, what needs the cloud, with accuracy and cost per 1,000 pages.

Show it to

Eywa Biotech, AVaTAR MedTech, NanoproX, Synthetic Ocular Restoration, SphereBio, Neocell, any pharma QA manager.

Preprint 2

"On-premise document intelligence for GxP laboratories: a calibrated evaluation of small vision-language models on synthetic batch records and SOPs". arXiv cs.CL/cs.DL, or JOSS for the software.

Phase-2 backlog · do not start before week 20

Instrument 4, clinical report automation (archetype E: Enteria, Microgenesis, PreGen, CASPR, Ubique). Higher ticket, regulated (HIPAA/GDPR/Ley 25.326), needs a partner clinic. Start only if a diagnostic company asks after seeing labdocs.

Scripts-to-product engineering for archetype F companies with software gaps (Geoprot, ArgenTag, Monte Caldera, Zavia). Off-thesis but lucrative: only as a paid engagement, never a portfolio project.

§6 · Theory catalog

What you study, why, and when

All free. Two 1.5-h theory blocks a week. Each piece names the practice it feeds; nothing is studied "for later".

Squares mark the weeks each piece is scheduled; the grey band is the science literacy thread, which runs all 20 weeks. Tick a piece when studied; click its name for source and payoff.

Deferred, and why: Karpathy and CS336 internals (learn them when an instrument needs them), MIT 6.874 (three lectures allowed inside the literacy thread), full DoE or biostatistics courses (the NIST handbook covers it), protein design (Zavia and Nanogrow's territory; you are not competing there).

§7 · Week-by-week plan

Twenty-one weeks, four gates

Tick lecture, practice and public output as you finish them. The calendar bends; the gates do not.

Weekly rhythm · 1.5-h blocks

MonCode1.5 h
TueTheory1.5 h
WedCode1.5 h · gap note goes out
ThuTheorylast 30 min: literacy or a founder's paper
FriCode1.5 h
SatCodefirst to fall in a bad week
SunPublish1 h: log, board, gap detector, TODO
GGitHubL-SunLinkedIn build logL-WedLinkedIn gap noteMMediumXarXiv / bioRxivZZenodo DOIV60-s video
§8 · Science literacy thread

Read their language before you sell to them

30–45 minutes on Thursdays, inside the theory hours. Four notebooks in 20 weeks: literacy, not instruments. They never get a release, a DOI or a feature.

§9 · Publishing engine

Every post carries one number or one admitted failure

Scheduled deliverables across the program. A dot fills when that week's public output is ticked in §7.
ChannelFormatCadenceReal objective
GitHub labloopThree instrument repos: README numbers, --fake mode, make report, Zenodo DOIMon/Wed/Fri/SatThe proof. The only thing a senior engineer or CTO actually checks
LinkedInSunday build log (A) + Wednesday gap note (E)2 per weekFounders, R&D heads, QA managers, fund partners, recruiters
MediumDeep dive, 1,500–2,500 words, one chart that carries the argumentWeeks 2, 8, 13, 18What people read before they write to you
arXiv / bioRxivPreprint with reproducible code and DOIWeeks 8, 18Credibility with scientists. arXiv needs an endorser for a first submission; ask a founder-scientist in week 6
ZenodoDOI per releaseWeeks 7, 13, 17Citable software. Free, instant
Video60 seconds, terminal only (C)Weeks 8, 13, 18Founders watch; they do not read READMEs
XCopy of the Sunday log, more technicalOptional, 5 minWhere the bioprocess-AI people live
Talk20 minutesProposed week 20Turns the program into one conversation with a room
  • The Wednesday note is derived from the week's work: 10 minutes, two paragraphs. If it takes 30, skip it.
  • Every post has one number or one admitted failure. Never "the ultimate guide".
  • Archetypes, never named weaknesses. Name a company only to praise or after they engaged.
  • Zero EY data, names, clients or comparisons. All code on your time and hardware, commits dated.

Templates


    

  







  

    §10 · Hardware, software and budget
    
A lab's AI on a USD 300 computer

  

  

    

      
ResourceRole in the plan

      
        
RTX 3050 6 GB + 32 GB RAM + RyzenAll training (GBM, small CNNs, PLS, BO) and local 3–7B models (Q4). ESM-2 650M in fp16. Enough for everything in weeks 1–20

        
Debian mini boxesThe on-prem story: Mosquitto, Node-RED, InfluxDB, Grafana, eLabFTW, Postgres, embeddings and a 1–2B model. Week 17 benchmarks them explicitly

        
Ollama / llama.cpp / LM StudioQwen2.5 7B Q4, Qwen2.5-VL 3B or Gemma 3 4B (vision), bge-m3 or nomic-embed

        
Claude, OpenAI, Gemini, OpenRouterCloud comparators in the accuracy/cost tables; Claude for schema mapping and judge; Gemini for long documents

        
NVIDIA free APILarge open models (Nemotron, Llama 70B) as a third comparator at zero cost

        
Google Colab freeOnly if ESMFold or a bigger model is needed in the literacy thread

        
Hugging FaceModels, and a dataset card for the synthetic batch records and SOPs (week 17)

        
ESP32 + DS18B20 optionalUSD 15. One real sensor in the demo makes the IoT conversation concrete

      
    

  

  

    
APIs, 20 weeks
150–200USD
Only weeks 5 and 15 extraction comparisons are expensive.

    
Hardware
0–15USD
One optional sensor.

    
Everything else
0USD
Free courses, free tiers, open data.

  







  

    §11 · Metrics and gates
    
The Sunday board, 30 seconds

    
Log this week's numbers every Sunday. Each week keeps its own board.

  

  

  

    

      
Founder conversations, cumulative

      

      
      
Targets: 5 by week 9, 15 by week 16, 30 by week 20. The strategy's leading indicator.

    

    

      

        

          
WeekGateIf missed

          
        

      

      

        
Vanity metrics, banned

        
starslikesfollowershours of videocourses completed

      

    

  

  
Honest expectation for week 20: evidence, one or two paid audits, a pipeline of conversations, and a clear view of which archetype pays fastest. The retainer or the job offer is the month 6–9 target.







  

    §12 · Ignore for now
    
Not in these twenty weeks

  

  
    
    
  • Vara as a flagship. Its harness and judge live on inside batchlens (week 5) and labdocs (week 15).
    • 
    
  • Instrument 4 (clinical reporting) before week 20, and any regulated clinical data at all in this phase.
    • 
    
  • Protein or molecule design. Zavia, Nanogrow, CellCo own that. You do literacy, not competition.
    • 
    
  • Building a LIMS/ELN product. eLabFTW exists. You add AI to it; you do not replace it.
    • 
    
  • A fourth instrument, a SaaS, a dashboard product, a marketplace. Module 23 in a lab coat.
    • 
    
  • Hourly consulting. Fixed price only in year one.
    • 
    
  • Wet-lab work, hardware beyond one sensor. You are the data-to-decision layer, not the lab.
    • 
    
  • Full courses: MIT 6.874 complete, CS336 complete, biostatistics MOOCs, cloud certifications.
    • 
    
  • Naming a company's weakness in public. Ever.
    • 
    
  • A second project. El Espejo stays at its current low-effort pace and does not count.
    • 
  







  

    §13 · Execution protocol
    
How each block runs

  

  

    
    
      
  • One block, one thing. Theory or code, never both in the same block.
    • 
      
  • Code first, theory when the code asks. Go to the lecture with a concrete question.
    • 
      
  • Tomorrow's TODO is the last line of every block, written in the repo.
    • 
      
  • Sunday hour is sacred: build log 20 min, board 5 min, NotebookLM gap detector 20 min, week TODO 15 min.
    • 
      
  • Publish even when the week was bad. "This week I did not advance; here is what blocked me" is a valid post.
    • 
      
  • Thursday's last 30 minutes go to literacy or one founder's paper. Reading their paper is the outreach.
    • 
      
  • Boredom is a signal. Three bored blocks on a topic means it is not connected to an instrument. Skip it.
    • 
      
  • Hygiene while employed: zero EY data or names, own hardware, dated commits, no comparisons with your employer.
    • 
    

    

      
Tomorrow's first block · 60–90 min

      
      
      
      
      
    

  







  

    §14 · The one thing to remember
    
In 20 weeks nobody in a lab will ask how many lectures you watched. A founder will ask: how many days does your instrument take out of my loop, how do you know, and where does it fail?

    
This document exists so that you can answer with a link, a number and an invoice.

  







  

    Appendix A · Full company map
    
86 companies from the GridX list

    
Grouped by primary fit. The gap column is a private hypothesis from the CSV, shortened, and never to be quoted in public, so it stays hidden until you switch it on.

  

  

  

    
    
    
  

  

    

      
CompanyFitAI/IoTWhat they doSecondary